MIT Technology Review 2020 07

cto er
LEADING WITH
INNOVATION
T Techno o Re e s
ann a a sh p e ent on e er n
techno o and trends oes on ne
Our world has hit an inflection point. Now is the time to reset,
rethink and rebuild. Join us online as our editors and experts
explore technology strategies for leadership in a changed world:
Cloud, AI, BioTech, Economy, Cybersecurity and more.
scr ers sa e th d sco nt code T at

Tech Tco Re ster
ETM20 JA20 Onsert TR D3.indd 1 6/4/20 9:43 AM

The innovation issue
The
innovation
issue
Volume 123 Jul/Aug USD $9.99
Number 4 2020 CAD $10.99
Pandemic.
Inequality.
Misinformation.
Unrest.
Technology has let us down.
Here’s how to make it
work for us again ...
and 35 young innovators
leading the charge
JA20_cover.indd 1 6/5/20 4:50 PM

October 20-22, 2020
LEADING
WITH
INNOVATION
Subscribers save 10% with discount code PRINTJA at
Tech Tco Re ster
ETM20 JA20 Spread D1.indd 1 6/4/20 9:46 AM

October 20-22, 2020
r or d has ore er chan ed

We need to plot a new path forward,
ead n th nno at on
MIT Technology Review’s premier online event

EmTech MIT brings together the leaders and
innovators who are redefining our future,
examining technology strategies for leadership
in a changed world: Cloud, AI, BioTech, Economy,
Cybersecurity and more.
Don’t be left out of the conversation. Come hear new

ideas, share your reactions, and be part of the reset
that is defining the next normal.
Subscribers save 10% with discount code PRINTJA at

Tech Tco Re ster
ETM20 JA20 Spread D1.indd 2 6/4/20 9:46 AM

02 From the editor
T
wo entrepreneurs share a goal: to help people
overcome opioid addiction. One, Zack Gray, has
an Ivy League education and $2.7 million in ven-
ture capital funding. The other, Nikki King, went
to the University of Kentucky and has a fraction
of that amount, cobbled together from grants,
donations, and Medicaid reimbursements.
But Gray’s investors will want their
money back some day. That means the only
people he can help are people who can pay.
This is the inexorable logic of venture capi-
tal, as Elizabeth MacBride writes (page 50):
it funnels money from people who have it
(customers) to people who have even more
of it (investors). Those who have none of
it have no say in one of the main driving
forces of American innovation.
The technology we have mirrors the
society we have, and specifically the way
power in that society is distributed. Those
who have power, be it through money, con-
nections, or other kinds of privilege, have
much more say in deciding which tech-
nologies get built and whom they benefit.
Such a system fails many people. Covid-
19 and, more recently, the protests in the
US sparked by the police officer who calmly
murdered the unarmed, unresisting George
Floyd in full view of cameras have made
this clearer than ever. The venture-capital- Gideon and direction, both of which have declined
Lichfield
driven tech boom of recent decades has not over the decades, as David Rotman (page
is editor
given the country much of the technology in chief of 6) and Ilan Gur (page 58) explain. More
and infrastructure it needs to fight a pan- MIT Technology muscular policy and regulation could also
Review.
demic. It has worsened economic inequal- help with the post-covid recovery, writes
ity, political polarization, and the spread of Nathan Schneider (page 48), by creating
misinformation. It has not reduced racial injustice: even though incentives and support for local entrepreneurs to build techno-
police brutality against black people has been documented count- logical solutions for their own communities.
less times on cell phones and police bodycams in the past few It’s hard to see much appetite for that kind of policymaking in
years, the death toll has stayed perfectly steady. Indeed, the US the current US government. For signs of hope, one might look to
has used technology to make racial oppression more systematic, Canada (page 42), where the tech hub of Toronto is trying (or so
as Charlton McIlwain writes (page 12). its boosters say, at any rate) to be a sort of gentler, kinder Silicon
The pandemic exacerbates these inequities. Not only are peo- Valley, driven less by rapacious capitalism and more by a concern
ple in some of the lowest-paid, most precarious jobs—delivery for technology’s social consequences.
drivers, supermarket cashiers, warehouse staff—at highest risk Look, too, at individual scientists, inventors, and entrepre-
of catching covid-19, but as Erika Hayasaki explains (page 64), neurs with ambitious, idealistic goals. As we do every year, we’ve
the crisis is likely to accelerate their replacement with robots and assembled a global and—importantly—diverse group of leading
other forms of automation. young innovators (page 15). We’ve also interviewed some of the
None of this is the fault of technology, but of a society that gives past years’ winners about what they’ve learned along their jour-
markets, and therefore the rich and powerful, too much say over neys (page 60). Their examples, we hope, can serve to inspire
which technologies are built and how they are used. This is not funders, policymakers, and other technologists with a reminder
IAN ALLEN
a call for socialism: free markets are essential to innovation. But of the good technology can do when it is directed at helping
America’s technology prowess owes much to government funding everyone—not just the moneyed and powerful.
JA20_editorial.indd 2 6/8/20 1:38 PM

W TO
NE
P
U
K
DS
00
AR
$5
AW
INNOVATE
TRANSFORM OUR FUTURE
R&D FUNDING PROGRAM
The National Reconnaissance Office Director’s Innovation Initiative (DII)
Program funds cutting-edge scientific research in a high-risk, high-payoff
environment to discover innovative concepts and creative ideas that transform
overhead intelligence capabilities and systems for future national security intelligence
needs. The program seeks the brightest minds and breakthrough technologies from
industry, academia, national laboratories, and U.S. government agencies.
Visit the website for Broad Agency Announcement and

Government Sources Sought Announcement requirements.
703.808.2769
www.nro.gov/Business-Innovation-Opportunities
Untitled-1 1 6/15/20 11:44 AM

04 The innovation issue
INNOVATORS THE STATE OF

UNDER 35 INNOVATION
Introduction 15 12 58
6 The list Of course technology How the US lost its way

Why tech In these chaotic times, there are still perpetuates racism. on innovation
didn’t save us many people trying to create a better It was designed that way. We’ve been using the same inno-
from covid-19 world. Here are 35 young innovators Black Americans have seen tech- vation playbook since World War
The pandemic working to solve our climate crisis, nology used to target them again II. A crisis shows us that it’s time
shows us that cure diseases, invent better batter- and again. Stopping it means look- for something new. By Ilan Gur
the US isn’t as ies, make AI more equitable, and ing at the problem differently.
innovative as we design better prosthetics, among By Charlton McIlwain 60
thought. Here’s many other things. Tales from the innovation

how to fix that. 42 trenches
By + Inventors . . . . . . . . . . . . . . . . . . . 16 Enter the narwhal We asked some previous winners
David Rotman + Pioneers . . . . . . . . . . . . . . . . . . 22 Can Canada save us from Silicon of our 35 Innovators Under 35
Valley’s worst impulses? award what they’ve learned about
+ Visionaries . . . . . . . . . . . . . . . . 28
By Brian Barth innovation since they won. Here’s
Fiction
+ Humanitarians . . . . . . . . . . . . . 34 what they told us.
70
+ Entrepreneurs . . . . . . . . . . . . . 38 48
Krishna A crisis is no excuse not 64
and Arjuna to regulate tech Serve and replace

By The Great Depression offers Could covid-19 accelerate the
Stephen Marche lessons for how to give ordinary robot takeover of human jobs?
people a say in the economic By Erika Hayasaki
76
recovery from covid-19.
How I used an By Nathan Schneider
algorithm to
help me write a 50
story Losing the winners’ game

By Venture capital has become
Stephen Marche extremely successful at creating
a certain kind of innovation—but
the pandemic has exposed its
COVER IMAGE: SHUTTERSTOCK
The back page broader failures.

78
By Elizabeth MacBride
The innovation
dilemma
JA20_Contents.indd 4 6/8/20 1:36 PM

Masthead 05
Editorial Corporate Consumer marketing MIT Technology Review

Editor in chief Chief executive officer and publisher Senior vice president, Insights and international
Gideon Lichfield Elizabeth Bramson-Boudreau marketing and consumer revenue
Vice president, Insights and international
Doreen Adger
Nicola Crepaldi
Executive editor Assistant to the CEO Director of analytics systems
Michael Reilly Katie McLean Director of custom content, US
Tom Russell Laurel Ruma
Editor at large Human resources manager Director of audience development
David Rotman James Wall Senior project manager
Rosemary Kelly Martha Leibs
News editor Manager of information technology Director of digital marketing
Niall Firth Colby Wheeler Content manager
Emily Baillieul Jason Sparapani
Managing editor Office manager Product marketing manager
Timothy Maher Linda Cardinal Client services manager,
Amanda Saeli
licensing and syndication
Commissioning editors Assistant consumer marketing manager Ted Hu
Bobbie Johnson Product development Caroline da Cunha
Director of custom content, international
Konstantin Kakaes
Chief technology officer Circulation and print production manager Claire Beatty
Senior editor, MIT News Tim Borton
Drake Martinet Director of business development, Asia
Alice Dragoon
Director of software engineering Email marketing specialist Marcus Ulvne
Senior editor, biomedicine Tuong-Chau Cai
Molly Frey
Antonio Regalado
Senior editor, energy
Head of product Board of directors
James Temple
Mariya Sitnova Advertising sales Martin A. Schmidt, Chair
Senior product designer Vice president, sales and Peter J. Caruso II, Esq.
Senior editor, digital culture
Jon Akland brand partnerships Whitney Espich
Abby Ohlheiser
Andrew Hendler Jerome I. Friedman
Director of analytics
Senior editor, cybersecurity andrew.hendler@technologyreview.com David Schmittlein
Michelle Bellettiere
Patrick Howell O’Neill 646-520-6981 Glen Shor
Senior project manager Alan Spoon
Senior editor, AI Executive director, brand partnerships
Allison Chase
Will Douglas Heaven Marii Sebahar
Senior editor, podcasts and
Senior software engineer marii@technologyreview.com Customer service and
Jason Lewicki 415-416-9140
live journalism subscription inquiries
Jennifer Strong Software engineer Executive director, brand partnerships
Jack Burns Kristin Ingram National
Senior reporters 800-877-5230
kristin.ingram@technologyreview.com
Tanya Basu (humans and technology)
415-509-1910 International
Karen Hao (AI) Events 903-636-1115
Senior director, brand partnerships
Reporters
Senior vice president, Whelan Mahoney Email
Charlotte Jee (news)
events and strategic partnerships whelan@technologyreview.com customer_service@
Neel Patel (space)
Amy Lammers 201-417-0928 mittechnologyreview.info
Copy chief
Director of event content Director, brand partnerships Web
Linda Lowenthal
and experiences Debbie Hanley www.technologyreview.com/
Social media editor Brian Bryson debbie.hanley@technologyreview.com customerservice
Benji Rosen 214-282-2727
Head of international and custom events MIT Records (alums only)
Editorial research manager Marcy Rizzo Director, brand partnerships 617-253-8270
Tate Ryan-Mosley Ian Keller
Event content producer ian.keller@technologyreview.com Reprints
Administrative assistant Erin Underwood 203-858-3396 techreview@wrightsmedia.com
Andrea Siegel 877-652-5295
Associate director of events Business development sales manager
Proofreader Nicole Silva Ken Collina Licensing and permissions
Barbara Wallraff ken.collina@technologyreview.com licensing@technologyreview.com
Event partnership coordinator
Madeleine Frasca 617-475-8004
Design Events associate Digital sales strategy manager MIT Technology Review
Bo Richardson Casey Sullivan
Chief creative officer casey.sullivan@technologyreview.com One Main Street
Eric Mongeon 617-475-8066 13th Floor
Art director Finance Media kit
Cambridge, MA 02142
617-475-8000
Emily Luong www.technologyreview.com/media
Finance director
The mission of MIT Technology Review
Marketing and events designer Enejda Xheblati
is to make technology a greater force for
Kyle Thomas Hemingway
General ledger manager good by bringing about better-informed,
Photo editor Olivia Male more conscious technology decisions
Stephanie Arnett through authoritative, influential, and
Accountant trustworthy journalism.
Letitia Trecartin
Technology Review, Inc., is an
independent nonprofit 501(c)(3)
corporation wholly owned by MIT;
the views expressed in our publications
and at our events are not always shared
by the Institute.
TR_masthead.indd 5 6/1/20 1:03 PM

T
echnology has failed the
US and much of the rest
of the world in its most
WHY TECH
important role: keeping
us alive and healthy. As I
write this, more than 380,000 are dead,
the global economy is in ruins, and the
covid-19 pandemic is still raging. In an
age of artificial intelligence, genomic
medicine, and self-driving cars, our
DIDN’T
most effective response to the outbreak
has been mass quarantines, a public
health technique borrowed from the
Middle Ages.
Nowhere was the technology failure
more obvious than in testing. Standard
tests for diseases like covid-19 use
polymerase chain reaction (PCR), a
more than 30-year-old chemistry tech-
SAVE US
nique routinely used in labs around the
world. Yet although scientists identified
and sequenced the new coronavirus
within weeks of its appearance in late
December—an essential step in cre-
ating a diagnostic—the US and other
countries stumbled in developing PCR
tests for general use. Incompetence
and a sclerotic bureaucracy at the US
Centers for Disease Control meant the
agency created a test that didn’t work
and then insisted for weeks that it was
the only one that could be used.
Meanwhile, the six-inch nasopha-
FROM COVID-19
ryngeal swabs needed to reach far up
a person’s nose to collect samples for
PCR testing were in short supply, as
were the chemical reagents necessary to
process the samples. In the critical early
weeks when the coronavirus could still
have been contained, many Americans,
The pandemic reveals even those seriously ill, couldn’t get
that the US is not nearly as tested for the deadly virus. Even four
innovative as we thought. months into the pandemic, the US still
isn’t equipped to do the massive and
Here’s how to fix that. frequent screening needed to safely
end a general lockdown.
Combined with the lack of testing,
a splintered and neglected system of
collecting public health data meant
epidemiologists and hospitals knew
By David Rotman
Illustration by Selman Design
JA20_Introduction.indd 6 6/5/20 4:38 PM

Where did all the money go?

US federal funding for R&D has fallen. That’s one cause of sluggish productivity growth. By Tate Ryan-Mosley
1 Federal funding has been dropping

for decades. (US R&D as % of GDP) 2 The US is lagging behind South Korea, Japan, and Germany, and China is
catching up. (R&D as % of GDP in 2000 vs. 2017)
2.8 2000 2017

2.6 Total
2.4
United States 2.62 2.81
2.2
2
EU 1.67 1.97
1.8 France 2.09 2.19
1.6 Federally funded
1.4 Germany 2.4 3.04
1.2
1 United Kingdom 1.63 1.66
0.8
0.6 Business funded China 0.89 2.15
0.4
0.2 Other nonfederally funded
Japan 2.91 3.21
0
South Korea 2.18 4.55
1 2 3 4 5
1955
1965
1975
1985
1995
2005
2015
NATIONAL SCIENCE FOUNDATION NATIONAL SCIENCE FOUNDATION
too little about the spread of the infection. A once-healthy innovation ecosystem in of government failure to support the pri-
In an age of big data in which companies the US, capable of identifying and creat- vate sector in doing so. Today, she says,
like Google and Amazon use all sorts of ing technologies essential to the country’s it feels as though she’s “living the book.”
personal information for their advertising welfare, has been eroding for decades. The US’s paralysis in the face of covid-
and shopping operations, health authori- Any country’s capacity to invent and 19 matters not only because it has already
ties were making decisions blind. then deploy the technologies it needs is doomed tens of thousands to an early death
It wasn’t only the lack of testing and data shaped by public funding and government and crippled the largest economy in the
that doomed so many people, of course. policies. In the US, public investment in world, but because it reveals a deep and
There weren’t enough ventilators or pro- manufacturing, new materials, and vac- fundamental flaw in how the nation thinks
tective masks, nor factories to make them. cines and diagnostics has not been a pri- about innovation.
“The pandemic has shone a bright light on ority, and there is almost no system of
just how much US manufacturing capa- government direction, financial backing, Building stuff we need
bilities have moved offshore,” says Erica or technical support for many critically Economists like to measure the impact
Fuchs, a manufacturing expert at Carnegie important new technologies. Without it, of innovation in terms of productivity
Mellon University. the country was caught flat-footed. growth, particularly “total factor produc-
Why couldn’t the US’s dominant tech Instead, as Henderson writes in her tivity”—the ability to get more output
industry and large biomedical sector pro- book Reimagining Capitalism, the US has, from the same inputs (such as labor and
vide these things? It’s tempting to simply over the last half-century, increasingly put capital). Productivity growth is what makes
blame the Trump administration’s inaction. its faith in free markets to create innova- advanced nations richer and more prosper-
Rebecca Henderson, an economist and tion. That approach has built a wealthy ous over the long run. For the US as well
management expert at Harvard, points Silicon Valley and giant tech firms that as most other rich countries, this measure
to a long history of the US government’s are the envy of entrepreneurs around the of innovation has been dismal for nearly
directing industry and innovation during world. But it has meant little investment two decades.
crises. Many companies, she says, were and support for critical areas such as manu- There are a lot of different ideas about
waiting for the administration to mobi- facturing and infrastructure—technologies why the innovation slowdown happened.
lize the effort and guide priorities. “I kept relevant to the country’s most basic needs. Perhaps the kinds of inventions that pre-
thinking, ‘Let’s focus the US thoughtful- Though written before covid-19 viously transformed the economy—like
ness on testing and we’ll get this.’ I kept emerged, Henderson’s book was pub- computers and the internet, or before that
waiting for it to happen,” she says. But it lished in mid-April, as the pandemic was the internal-combustion engine—stopped
never did: “There is simply a vacuum.” surging in many parts of the US. In it, she coming along. Or perhaps we just hav-
But Henderson and other experts who describes the role business can play in tack- en’t yet learned how to use the newest
study innovation point to a problem deeper ling big problems like climate change and technologies, like artificial intelligence,
than the lack of government intervention. inequality, but she also documents decades to improve productivity in many sectors.

Introduction 09
3 Spending on basic research has

been nearly flat. ($ million) 4 Total factor productivity (TFP)
is sluggish. (Growth rate, %) 5 And manufacturing TFP has
collapsed. (Growth rate, %)
600,000 ■ Development 1.6 8

■ Applied
■ Basic 1.4
500,000
6
1.2
400,000
1
4
300,000 0.8
2
0.6
200,000
0.4
0
100,000
0.2
0 0 -2
2000–
2010
2011
2013
2014
2015
2016
2017
2018
2005–
1990
2000
2005
2010
2015
2012
1995
1990–
2010–
1985–
1995–
2015–
2004
2009
1999
2018
1994
2014
1989
NATIONAL SCIENCE FOUNDATION OECD US BUREAU OF LABOR STATISTICS
But one likely factor is that governments Andreessen decried the US’s inability to making many things, including solar pan-
in many countries have significantly cut “build” and produce needed supplies like els and advanced batteries—and, it now
investments in technology since the 1980s. masks, claiming that “we chose not to turns out, swabs and diagnostic tests too.
Government-funded R&D in the US, have the mechanisms, the factories, the No country should aim to make every-
says John Van Reenen, an economist at systems to make these things.” The accusa- thing, says Fuchs, but “the US needs to
MIT, has dropped from 1.8% of GDP in tion resonated with many: the US, where develop the capacity to identify the tech-
the mid-1960s, when it was at its peak, to manufacturing has deteriorated, seemed nologies—as well as the physical and
0.7% now (chart 1). Governments tend to unable to churn out things like masks and human resources—that are critical for
fund high-risk research that companies ventilators, while countries with strong and national, economic, and health security,
can’t afford, and it’s out of such research innovative manufacturing sectors, such as and to invest strategically in those tech-
that radical new technologies often arise. China, Japan, Taiwan, and Germany, have nologies and assets.”
The problem with letting private invest- fared far better. Regardless of where products are made,
ment alone drive innovation is that the But Andreessen is wrong to portray Fuchs says, manufacturers need more coor-
money is skewed toward the most lucra- the unwillingness to build as a deliberate dination and flexibility in global supply
tive markets. The biggest practical uses of choice. And the country’s ability to make chains, in part so they aren’t tied to a few
AI have been to optimize things like web stuff isn’t something that can be quickly sources of production. That quickly became
search, ad targeting, speech and face rec- revved up. The decline of US manufactur- evident in the pandemic; for example, US
ognition, and retail sales. Pharmaceutical ing has been caused by years of financial mask makers scrambled to procure the lim-
research has largely targeted the search market pressures, government indiffer- ited supply of melt-blown fiber required to
for new blockbuster drugs. Vaccines and ence, and competition from low-wage make the N95 masks that protect against
diagnostic testing, so desperately needed economies. the virus.
now, are less lucrative. More government In the US, manufacturing jobs dropped The problem was made worse because
money might have boosted those pursuits. by almost a third between 2000 and manufacturers keep inventories razor-thin
Nor is it enough to invent new tech- 2010 and have barely recovered since. to save money, often relying on timely
nologies: public support is also vital for Manufacturing productivity has been par- shipments from a sole provider. “The great
helping companies adopt them. That’s ticularly poor in recent years (chart 5). lesson from the pandemic,” says Suzanne
especially true in large, slow-moving sec- What has been lost is not only jobs but Berger, a political scientist at MIT and an
tors of the economy such as health care also the knowledge embedded in a strong expert on advanced manufacturing, is “how
and manufacturing—precisely where the manufacturing base, and with it the ability we traded resilience for low-cost and just-
country’s crippled capabilities have been to create new products and find advanced in-time production.”
most evident during the pandemic. and flexible ways of making them. Over the Berger says the government should
In a widely circulated blog post, inter- years, the country ceded to China and other encourage a more flexible manufacturing
net pioneer and Silicon Valley icon Marc countries the expertise in competitively sector and support domestic production

by investing in workforce training, basic technologies crucial to handling the cur- The thing to note about all these pro-
and applied research, and facilities like rent crisis, such as tests and vaccines, and posals is that they are aimed at both short-
the advanced manufacturing institutes in new jobs and economic revival. Many and long-term problems: they are calling
that were created in the early 2010s to of the jobs created will be for scientists, for an immediate ramp-up of public invest-
provide companies with access to the lat- Johnson acknowledges, but many will also ment in technology, but also for a bigger
est production technologies. “We need to go to trained technicians and others whose government role in guiding the direc-
support manufacturing not only [to make] work is needed to build and maintain an tion of technologists’ work. The key will
critical products like masks and respira- enlarged scientific infrastructure. be to spend at least some of the cash in
tors but to recognize that the connection This matters especially, he says, because the gigantic US fiscal stimulus bills not
between manufacturing and innovation with an administration that is pulling back just on juicing the economy but on reviv-
is critical for productivity growth and, out from globalization and with consumer ing innovation in neglected sectors like
of increases in productivity, for economic spending weak, innovation will be one advanced manufacturing and boosting
growth,” she says. of the few options for driving economic the development of promising areas like
The good news is that the US has had growth. “Scientific investment needs to be AI. “We’re going to be spending a great
this discussion during previous crises. The a strategic priority again,” says Johnson. deal of money, so can we use this in a
playbook exists. “We’ve lost that. It has become a residual. productive way? Without diminishing the
That’s got to stop.” enormous suffering that has happened,
Declaring war on the virus Johnson is not alone. In the middle of can we use this as a wake-up call?” asks
In June 1940, Vannevar Bush, then the May, a bipartisan group of congressmen Harvard’s Henderson.
director of the Carnegie Institution for proposed what they called the Endless “Historically, it has been done a bunch
Science in Washington, DC, went to the Frontier Act to expand funding for “the of times,” she says. Besides the World War
White House to meet President Franklin discovery, creation, and commercialization II effort, examples include Sematech, the
D. Roosevelt. The war was under way in of technology fields of the future.” They 1980s consortium that revived the ailing
Europe, and Roosevelt knew the US would argued that the US was “inadequately pre- US semiconductor industry in the face of
soon be drawn into it. As Simon Johnson pared” for covid-19 and that the pandemic Japan’s increasing dominance, by sharing
and Jonathan Gruber, both economists “exposed the consequences of a long-term technological innovations and boosting
at MIT, write in their recent book Jump- failure” to invest in scientific research. The investment in the sector.
Starting America, the country was woefully legislators called for $100 billion over five Can we do it again? Henderson says
unprepared, barely able to make a tank. years to support a “technology directorate” she is “hopeful, though not necessarily
Bush presented the president with a plan that would fund AI, robotics, automation, optimistic.”
to gear up the war effort, led by scientists advanced manufacturing, and other criti- The test of the country’s innovation
and engineers. That gave rise to the National cal technologies. system will be whether over the coming
Defense Research Committee (NDRC); Around the same time, a pair of econo- months it can invent vaccines, treatments,
during the war, Bush directed some 30,000 mists, Northwestern’s Ben Jones and MIT’s and tests, and then produce them at the
people, including 6,000 scientists, to steer Pierre Azoulay, published an article in massive scale needed to defeat covid-19.
the country’s technological development. Science calling for a massive government- “The problem hasn’t gone away,” says
The inventions that resulted are well led “Pandemic R&D Program” to fund and CMU’s Fuchs. “The global pandemic will
known, from radar to the atomic bomb. But coordinate work in everything from vac- be a fact of life—the next 15 months, 30
as Johnson and Gruber write, the invest- cines to materials science. The potential months—and offers an incredible oppor-
ment in science and engineering continued economic and health benefits are so large, tunity for us to rethink the resiliency of our
well after the war ended. “The major—and Jones argues, that even huge investments supply chains, our domestic manufacturing
now mostly forgotten—lesson of the post- to accelerate vaccine development and capacity, and the innovation around it.”
1945 period is that modern private enter- other technologies will pay for themselves. It will also take some rethinking of how
prise proves much more effective when Vannevar Bush’s approach during the the US uses AI and other new technologies
government provides strong underlying war tells us it’s possible, though the fund- to address urgent problems. But for that
support for basic and applied science and ing needs to be substantial, says Jones. But to happen, the government has to take on
for the commercialization of the resulting increased funding is just part of what is a leading role in directing innovation to
innovations,” they write. required, he says. The initiative will need meet the public’s most pressing needs.
A similar push to ramp up government a central authority like Bush’s NDRC to That doesn’t sound like the government
investment in science and technology “is identify a varied portfolio of new technol- the US has now.
clearly what we need now,” says Johnson. ogies to support—a function that is miss- David Rotman is editor at large of
It could have immediate payoffs both in ing from current efforts to tackle covid-19. MIT Technology Review

Share your opinion
about today’s
top tech trends.
Join our Global Panel.
As a member of MIT Technology Review’s fast-growing
group of executives and innovators, you can take part
in surveys and weigh in on the impact of AI, blockchain,
and other technologies—plus get in-depth news and
analysis to inform business strategy.
Apply today and get our latest research report free:

insights.technologyreview.com
MA20 Global Panel Ad D1.indd 1 1/31/20 10:48 AM

OF COURSE
So the question we have to confront is
whether we will continue to design and
deploy tools that serve the interests of
racism and white supremacy.
TECHNOLOGY
Of course, it’s not a new question at all.
Uncivil rights
In 1960, Democratic Party leaders con-
PERPETUATES RACISM.
fronted their own problem: How could
their presidential candidate, John F.
Kennedy, shore up waning support from
black people and other racial minorities?
IT WAS DESIGNED
An enterprising political scientist at
MIT, Ithiel de Sola Pool, approached them
with a solution. He would gather voter data
from earlier presidential elections, feed
THAT WAY.
it into a new digital processing machine,
develop an algorithm to model voting
behavior, predict what policy positions
would lead to the most favorable results,
and then advise the Kennedy campaign to
act accordingly. Pool started a new com-
Black Americans have seen technology used to target them again pany, the Simulmatics Corporation, and
and again. Stopping it means looking at the problem differently. executed his plan. He succeeded, Kennedy
By Charlton McIlwain
was elected, and the results showcased
the power of this new method of predic-
T
tive modeling.
Racial tension escalated through-
oday the United States crumbles out the 1960s. Then came the long, hot
under the weight of two pandemics: summer of 1967. Cities across the nation
coronavirus and police brutality. burned, from Birmingham, Alabama, to
Both wreak physical and psycho- Rochester, New York, to Minneapolis,
Minnesota, and many more in between.
logical violence. Both dispropor-
Black Americans protested the oppres-
tionately kill and debilitate black
sion and discrimination they faced at
and brown people. And both are ani-
the hands of America’s criminal justice
mated by technology that we design, system. But President Johnson called it
repurpose, and deploy—whether “civil disorder,” and formed the Kerner
it’s contact tracing, facial recognition, or social media. Charlton
Commission to understand the causes of
McIlwain
We often call on technology to help solve problems. is a professor “ghetto riots.” The commission called on
But when society defines, frames, and represents peo- of media, cul-
Simulmatics.
ture, and commu-
ple of color as “the problem,” those solutions often do nication at New As part of a DARPA project aimed at
more harm than good. We’ve designed facial recognition York University turning the tide of the Vietnam War, Pool’s
and author of
technologies that target criminal suspects on the basis Black Software: company had been hard at work preparing
of skin color. We’ve trained automated risk profiling The Internet & a massive propaganda and psychological
Racial Justice,
systems that disproportionately identify Latinx peo- From the AfroNet
campaign against the Viet Cong. President
ple as illegal immigrants. We’ve devised credit scoring to Black Lives Johnson was eager to deploy Simulmatics’
Matter behavioral influence technology to quell
algorithms that disproportionately identify black people
the nation’s domestic threat, not just its
COURTESY PHOTO
as risks and prevent them from buying homes, getting

foreign enemies. Under the guise of what
loans, or finding jobs.
it called a “media study,” Simulmatics built
a team for what amounted to a large-scale
JA20_McIlwain-essay.indd 12 6/5/20 1:15 PM

Essay 13
jail, or secretly “disappear” them. Instead,

by the end of the 1960s, this kind of data
had helped create what came to be known
as “criminal justice information sys-
tems.” These then proliferated, laying the
foundation for racial profiling, predictive
policing, and racially targeted surveillance.
They left behind a legacy that includes
millions of black and brown women and
men incarcerated.
Reframing the problem

Blackness and black people. Both per-
sist as our nation’s—dare I say even our
world’s—problem. When contact tracing
first cropped up at the beginning of the
covid-19 pandemic, it was easy to see it
as a necessary but benign health sur-
veillance tool. The coronavirus was the
world’s problem, and new surveillance
technologies like exposure notification,
temperature monitoring, and threat map-
ping began emerging to help address it.
But something both curious and tragic
happened. It was discovered that black
people, Latinx people, and indigenous
populations were disproportionately
infected and affected. Suddenly, we—
When the rioting started, black people those black and brown people—also
became a national problem; we dispropor-
were again seen as a threat to law and tionately threatened to spread the virus.
order, a threat to a system that perpetuates That was compounded when the tragic
white racial power. murder of George Floyd by a white police

officer sent thousands of protesters into
the streets. When the looting and rioting
started, we were again seen as a threat to
law and order, a threat to a system that
surveillance campaign in the “riot-affected before and during the protests, and how perpetuates white racial power. It makes
areas” that captured the nation’s attention they prepared for the aftermath. They col- you wonder how long it will take for law
that summer of 1967. lected data on toll booth usage, gas station enforcement to deploy those technologies
Three-member teams went into sales, and bus routes. They gained entry first designed to fight covid-19 to quell
areas where riots had taken place. They to these communities under the pretense the threat that black people supposedly
identified and interviewed strategically of trying to understand how news media pose to the nation’s safety.
important black people. They followed up supposedly inflamed “riots.” But Johnson If we don’t want our technology to
ILLUSTRATION SOURCE: GETTY IMAGES
to identify and interview other black res- and the nation’s political leaders were try- be used to perpetuate racism, then we
idents, in every venue from barbershops ing to solve a problem. They aimed to use must make sure that we don’t conflate
to churches. They asked people what they the information that Simulmatics collected social problems like crime or violence
thought about the news media’s coverage to trace information flow during protests or disease with black and brown people.
of the “riots.” to identify influencers and decapitate the When we do that, we risk turning those
But they collected data on much more, protests’ leadership. people into the problems that we deploy
too: how people moved in and around the They didn’t accomplish this directly. our technology to solve, the threat we
city during the unrest, who they talked to They did not murder people, put people in design it to eradicate.
JA20_McIlwain-essay.indd 13 6/5/20 11:19 AM

ATTENTION
We launched
a new podcast
just for you.
MAGAZINE
Deep Tech brings you
closer to the people,
places, and ideas
featured in the pages
SUBSCRIBER!
of MIT Technology
Review magazine.
Listen at
technologyreview.com/DeepTech
MJ20 Podcasts D2.indd 1 3/31/20 11:01 AM

35 Innovators
Slug here
Under 35 15
35 Innovators
Under 35
In chaotic times it can be reassuring JUDGES |—————————————— Oren Etzioni Hao Li
to see so many people working toward CEO, Allen Institute for AI; CEO, Pinscreen; Associate Pro-
Professor of Computer Sci- fessor, University of Southern
a better world. That’s true for medical Nora Ayanian
ence, University of Washington California; Director, USC Insti-
Associate Professor, Andrew
professionals fighting a pandemic and tute for Creative Technologies
and Erna Viterbi Early Career
for ordinary citizens fighting for social Chair, Computer Science, Uni-
David Fattal
Founder and CEO, Leia Inc. Nicole Paulk
justice. And it’s true for those among us versity of Southern California
Assistant Professor, University
striving to employ technology to address Chelsea Finn of California, San Francisco
Burcin Becerik-Gerber
those problems and many others. Professor of Civil and Environ-
Assistant Professor of Com-
puter Science and Electrical Carmichael Roberts
The 35 young innovators in these pages mental Engineering and Direc-
Engineering, Stanford Uni- Founder, Material Impact
aren’t all working to fight a pandemic tor of the Center for Intelligent
versity
Environments, University of
(though some are: see Omar Abudayyeh, Southern California
John Rogers
page 16, and Andreas Puschnik, page 27), Javier Garcia Martinez Simpson/Querrey Professor
Professor of Inorganic Chem- of Materials Science and Engi-
and they’re not all looking to remedy social David Berry
istry, University of Alicante, neering, Biomedical Engineer-
CEO, Integral Health; General
injustices (though some are: see Inioluwa Partner, Flagship Pioneering
Spain; President-elect, IUPAC ing, and Neurological Surgery,
Deborah Raji, page 28, and Mohamed Northwestern University
Julia Greer
Dhaouafi, page 37). But even those who Ed Boyden
Mettler Professor of Materials, Rachel Sheinbein
Y. Eva Tan Professor in Neuro-
aren’t tackling those specific problems Mechanics, and Medical Engi- Venture Partner, Lemnos
technology, MIT
are seeking ways to use technology to neering, Caltech
Cyrus Wadia
help people. They’re trying to solve our Yet-Ming Chiang
Zhen Gu Head of Sustainable Product,
Kyocera Professor, Materials
climate crisis, find a cure for Parkinson’s, Professor, University of Amazon
Science and Engineering, MIT
or make drinking water available to those California, Los Angeles
Jennifer West
who are desperate for it. James Collins
Ilan Gur Fitzpatrick Family Professor of
This contest generates more than 500 Termeer Professor, MIT
Founder and CEO, Activate Engineering, Duke University
nominations each year. The editors then John Dabiri
Ayanna Howard Jackie Ying
face the task of picking 100 semifinalists Centennial Professor of Aero-
Linda J. and Mark C. Smith A*STAR Senior Fellow,
to put in front of our 25 judges, who have nautics and Mechanical Engi-
Professor and Chair, School NanoBio Lab
neering, Caltech
expertise in artificial intelligence, biotech- of Interactive Computing,
nology, software, energy, materials, and Georgia Tech Ben Zhao
Gozde Durmus
Neubauer Professor of Computer
so on. With the invaluable help of these Assistant Professor, Stanford
Science, University of Chicago
University
rankings, the editors pick the final list of
35. —The editors
JA20_35-introduction.indd 15 6/8/20 1:02 PM

16 35 Innovators Under 35
enzyme, Cas9. The tool, it turned out, was

easy to use and worked in many species.
Biotech startups began racing to treat
genetic disease in humans. Gene-edited
human twins were even born in China.
Inventors
During what he calls the “Cas9 craze,”
Abudayyeh was drafted into a less visible
avenue of research: the effort to discover
and characterize novel CRISPR enzymes.
Soon the list was growing, and
Abudayyeh and colleagues were demon-
Their innovations point toward a strating what the new editors could do.
There would be Cpf1, also known as
future with new types of batteries, Cas12a, and then Cas12b. But one called
Cas13, discovered literally under our noses
solar panels, and microchips. (it’s part of a human oral bacterium called
Leptotrichia shahii), was special. Instead
of cutting DNA, the enzyme could instead
target RNA, the genetic messenger mol-
ecule inside of cells, which is also the
OMAR ABUDAYYEH
primary genetic material of many viruses,
including the coronavirus.
It was a totally new way to edit. What
MIT hadn’t changed was Abudayyeh’s close
around 2% of Americans had been tested and ongoing collaboration with fellow
Age 30 | Country of birth: US
for covid-19. Some economists say the gene editor Jonathan Gootenberg. The
country needs to test that many people pair first met as MIT undergrads and
He’s working to use CRISPR as
every day to reopen with confidence. then worked together in the busy lab
a covid-19 test that you could take
at home. That’s why, since January, Abudayyeh of CRISPR pioneer Feng Zhang (who
and his colleagues have been trying to made our list of 35 innovators in 2013) at
CRISPR has been called the discovery of forge CRISPR into an at-home test for the Broad Institute. They’ve written 28
the century for its potential to change bio- the virus. The basic chemistry is simple papers together, and in 2017 they were
medical research and treatment of genetic enough, they think, to create an easy- hired to establish a joint lab at the MIT
diseases. But it was Omar Abudayyeh who to-use test that you could give yourself McGovern Institute, which they chris-
helped turn the gene-editing tool into a before heading to work, or maybe take tened the “AbuGoot Lab.”
diagnostic test, one that might help slow at an airport gate before catching a flight. “We joke that it’s a scientific bro-
down the covid-19 pandemic. If they succeed, virus testing could mance that just keeps on going,” says
Seizing on the precise gene-finding happen anywhere, anytime, and the Abudayyeh, who reckons he’s the more
mechanism, in 2016 Abudayyeh, along gene-editing revolution would reach practical of the two, while Gootenberg is
with Jonathan Gootenberg and other col- directly into people’s homes and lives more mathematical. “Our brains haven’t
leagues at MIT, forged CRISPR into a tool for the first time. quite merged, but it’s close.”
to spot cancer mutations, bacteria, and And they needed two heads to under-
mosquito-borne viruses like Zika. Soon, stand the new RNA editor, Cas13. The
there was a spinout startup company Here’s how it works enzyme turned out to have a bizarre “col-
called Sherlock Biosciences, $49 million lateral effect.” Not only did it cut specific
in funding, and newspaper stories about The CRISPR revolution began with dis- RNA strands, but once it got going, it
CRISPR’s “new capabilities.” coveries, in the early 2000s, that bacteria would furiously chop up and degrade
Then came covid-19. Genetic tests to had evolved a way to chop up marauding any RNA in its path. “The mechanism
spot the pathogen were in desperately phage viruses. CRISPR, whose name is an was insane and very confusing at first,”
short supply in the US, with the work- acronym for this natural biological inven- says Abudayyeh. “We think it’s part of
horse technology, PCR, floundering. By tion, can spot unique sequences of DNA a cell-suicide mechanism”—a natural
early May, three months into the outbreak, letters and cleave them with a cutting self-destruct device in bacteria attacked
JA20_35-inventors.indd 16 6/4/20 2:03 PM

Inventors 17
Portrait by David Vintiner
by a virus. “When it activates, it shuts was first floated by scientists from the Great idea, but on its own, Cas13
down everything in the cell.” rival laboratory of Jennifer Doudna at wasn’t sensitive enough to create a test.
The indiscriminate cutting, though, the University of California, Berkeley. So Abudayyeh and Gootenberg got help
meant Cas13 wasn’t a great editor on its There a team proposed that indiscrim- from MIT professor Jim Collins, who
own. “It was kind of disappointing, but we inate cutting could serve as a detection showed them how to add a preamplifica-
came from an engineering background, mechanism. In short, if the enzyme found tion step, or a way to copy and multiply the
so we asked what it is good for,” says a match in a test tube—a piece of RNA RNA before testing for a match. By 2017,
Abudayyeh. Maybe they could blow up belonging to a virus, say—the collateral the group was showing off a complete
RNA in a cancer cell, bringing it to a halt? cutting could be used to sever special CRISPR diagnostic system called Sherlock
The idea that the collateral damage RNA that, when broken, would set off a that could locate unique mutations that
could turn CRISPR into a lab diagnostic visible fluorescent signal. cause cancer or flag the presence of

“I think our goal right now is to have it

ANASTASIA VOLKOVA
ready for the fall. For when the second Flurosat
wave comes.” Age 28 | Country of birth: Ukraine
Her platform uses remote sensing

and other techniques to monitor
crop health—helping farmers
focus their efforts where they’re
bacteria, or even the Zika virus. And campus, Abudayyeh, Gootenberg, and
most needed.
it was highly accurate. Imagine being Zhang set out to simplify the technol-
able to pick out one person’s face from ogy. They reasoned that if they could
the population of 100 million Earths. eliminate some of the fluid mixing If there’s one thing that frustrates
That’s the equivalent of what Sherlock steps, the test could be used in work- Anastasia Volkova, it’s inefficiency. So
could do in sorting through RNA. places, in pharmacies, or even at home. when she realized she could combine
Sherlock soon had competition from It didn’t need repeated heating and remote sensing data with scientific mod-
the Berkeley team, which started its cooling, as PCR does. And the readout eling to improve crop yields, reduce the
own CRISPR diagnostics company, was easy to understand: just colored use of agricultural chemicals, and make
Mammoth Biosciences. One result: a bars on a paper strip, like a pregnancy better use of water, she knew she’d found
tangle of competing diagnostic pat- test. “Our vision is really testing that her life’s work. It didn’t matter that she
ents that is reminiscent of the bruising, can be done at home,” says Abudayyeh. was still pursuing her doctorate in aero-
costly fight between the two institutions “So how can we push this so it’s fewer space at Sydney University or that she
over the original CRISPR inventions. steps, simple, and cheap?” would need to single-handedly raise more
Abudayyeh shrugs: “It’s more exciting Right now, so-called point-of-care than $5 million in startup money: Volkova,
when you have more than one group diagnostic tests do exist, but they need the daughter of a self-taught botanist and
working on it. And it’s better for CRISPR to be run on machines that cost thou- the goddaughter of a successful farmer,
diagnostics that it’s not just one com- sands of dollars. One device, ID NOW, wanted to fix what she thought was wrong
pany trying to peddle a technology.” which is sold by Abbott, returns corona- with large-scale farming.
He’s right: reaching the market is virus results in 15 minutes and is used Her resulting venture, Flurosat, uses
the hard part. That’s because diagnostic by the White House to screen visitors imaging sensors on satellites, planes,
testing is a business of giant companies, meeting with President Donald Trump. and drones to detect when crops are in
big machines, and centralized labs. It But the machine that processes the test trouble long before their distress is dis-
can take a hundred million dollars to costs thousands to buy. Abudayyeh cernible to the naked eye. Like humans,
develop a test that sells for $45. “Not says CRISPR home tests might cost $6 plants spike a fever when they’re sick.
for the faint of heart” is how venture each and only use simple equipment. They also heat up in response to pests or
capitalist Bruce Booth once described By May, the researchers had created because they’re not getting the nutrition
the business. By late 2019, Sherlock, a a simplified version and launched a or water they need. Flurosat uses multi-
company Abudayyeh cofounded, was website to share the new chemistry, spectral and thermal cameras to record
still edging the CRISPR-based tests which they showed could spot the these changes and AI to calibrate crop
toward the market. coronavirus in swabs from patients. models. Comparing a real crop with its
But then the pandemic exploded They are working with a design firm to digital twin then enables Volkova and her
out of China and changed everything. create a prototype of a plastic cartridge team to make real-time recommendations
When the shortage of tests in the US to hold and mix the test ingredients. to agronomists and farm managers about
became clear, the Food and Drug So has Abudayyeh tested himself? He what their yields need to thrive.
Administration started giving emer- hasn’t. “It’s tempting to spit in the This kind of monitoring and support
gency approvals to makers of dozens tube,” he says. “But it’s also a scary could reduce the overuse of nitrogen,
of tests, allowing them into the mar- thing to do.” pesticides, and herbicides and optimize
ket immediately. In May, Sherlock Pretty soon, though, people around irrigation. —Kathryn Miles
Biosciences won US authorization to the world may be having such “Do I or
perform a version of the CRISPR test don’t I have it?” moments regularly, or
that had to be done in a lab, although at least that’s the hope.
at press time no one had yet used it The work is “not final,” Abudayyeh
on a patient. says. “Final is a simple device you can
spit in. But this is the version of the
chemistry that would work for the
A home test home. I think our goal right now is to
COURTESY PHOTO
have it ready for the fall. For when the

Still, it wasn’t easy enough for someone second wave comes.”
without training to use. Back on MIT’s —Antonio Regalado

Inventors 19
LEILA PIRHAJI
ReviveMed
Age 34 | Country of birth: Iran
She developed an AI-based system that

can identify more small molecules in a patient’s
body, faster than ever before.
MANUEL LE GALLO
Leila Pirhaji built an AI-based the body,” she says. “They blood sugars and choles-
tool for measuring tiny mol- are involved in our metab- terol to obscure molecules
ecules in the body called olism and are downstream that appear in significant
metabolites, and her work from DNA, so they show the numbers only when some- IBM RESEARCH
could help us better detect effects of both our genes one is sick.
and treat diseases. “There and lifestyle.” Such metabo- The problem is that Age 34 | Country of birth: Canada
are 100,000 metabolites in lites include everything from measuring and identifying
metabolites is expensive He uses novel computer designs
and time consuming, and to make AI less power hungry.
fewer than 5% of metab-
CHRISTINA BOVILLE olites in a patient can be

identified using common
Training a typical natural-language pro-
cessor requires so much computing power
Aralez Bio
technologies. that it emits as much carbon as the life
So Pirhaji developed a span of five American cars. Training an
Age 32 | Country of birth: US platform that uses machine image recognition model releases as much
learning to do it much more energy as a typical home puts out in two
She modifies enzymes to enable production quickly. First she built a huge
weeks—and it’s something that leading
of new compounds for industry. database of all known infor-
tech companies do multiple times a day.
mation about existing metab-
Christina Boville helped design a pro- Much of the energy use in modern
olites and how they interact
cess that improves on biology’s way of with various proteins and computing comes from the fact that data
controlling chemical reactions. She other molecules. Then her needs to be constantly transferred back
starts with natural enzymes—pro- team collected tissue and and forth between memory and the pro-
teins that enable chemical reactions blood samples from patients cessor. Manuel Le Gallo is working with
in living cells—and engineers them with known diseases, and a research team at IBM that’s building
to produce useful chemicals that don’t exist in nature. measured the metabolites. technology to enable new kinds of com-
The approach can reduce manufacturing times for Her platform was able puting architecture that aims to be faster
compounds used in the pharmaceutical industry from to analyze the data, under- and more energy efficient but still highly
stand the complex connec-
months to days, shrink waste by up to 99%, and cut precise.
tions between diseases and
energy consumption in half. Le Gallo’s team developed a system
metabolites, and use this
In 2019, Boville cofounded Aralez Bio with David that uses memory itself to process data,
information to discover new
Romney and Frances Arnold, who won a Nobel Prize in drugs. When she tested it in and his team’s early work has shown they
2018 for a new way of creating enzymes called directed a mouse with Huntington’s can achieve both precision and huge
evolution. Boville’s process creates chemicals known as disease during her PhD at energy savings. The team recently com-
non-canonical amino acids (ncAAs), which are used in pleted a process using just 1% as much
COURTESY PHOTOS (PIRHAJI, BOVILLE); SAMUEL TRÜMPY (LE GALLO)
MIT, her team learned new

making 12% of the 200 best-selling medicines, includ- mechanisms for the disease energy as when the same process was
ing those for migraines and diabetes, and are also used and found new potential performed with conventional methods.
in agriculture. “Nature was built using 20 amino acids, ways of treating it. As companies from the financial sec-
and now our enzymes can make hundreds more,” she As CEO of ReviveMed, tor to life sciences constantly train their
Pirhaji is focusing on liver,
says. Drug ingredients “normally take five to 10 steps AI models to improve them, their energy
immune, inflammatory, and
to make,” she adds, “but we can do it in a single step.” needs will balloon. “What will change is
other diseases. Using her
Aralez Bio was recently approached by a pharma- we will be able to change models faster
platform, the startup part-
ceutical company to produce ncAAs that had taken ners with major pharmaceu- and more energy efficiently, which will
the company nine months to make with conventional tical companies to match definitely reduce the carbon footprint
methods. Boville’s enzymes now makes the same com- existing medicines to new and energy spent training those models,”
pound overnight. —Russ Juskalian treatments and find new Le Gallo says. —Patrick Howell O’Neill
targets for future drugs.
—Russ Juskalian

RANDALL JEFFREY PLATT
ETH Zurich
Her goal is to Age 32 | Country of birth: US
give anyone His recording tool provides a video

with an idea of genes turning on or off.
the means
to efficiently Randall Platt has created a way to record
translate it molecular events in a cell across time—a

technology that has the potential to trans-
into physical form our understanding of a number of
reality. important biological processes.
Currently, for instance, one of the best
tools available to understand the molec-
ular processes that occur during embry-
onic development or immune responses
to cancer is RNA-seq, a technique that
allows biologists to develop a snapshot of
how genes are being expressed—which
ones are being turned on or off—at a sin-
gle moment in time. But while RNA-seq
provides a snapshot, Platt’s tool could
potentially be used to record the equiv-
NADYA PEEK alent of a brief video, capturing gene

expression over time and thus providing
University of Washington a much richer picture of, say, an embryo’s
development.
Age 34 | Country of birth: US “At the core of all of
biology and biomedicine
She builds novel modular machines that can do just about anything is looking at transitions
you can imagine. in systems—whether
it be a stem cell that
Nadya Peek began tinkering with their creativity: they’ve made T-shirt- develops into a neuron
machines out of stubbornness. designing machines and cocktail-mixing or a healthy neuron that develops into
As an undergraduate, when she col- machines, 3D printers, and chemistry a degenerative neuron,” he says. “How
laborated with artists on their instal- pipetting machines. The machines are people approach this problem today is
lations, she often ran into limitations often no larger than a desktop and can
they perform time-point experiments
with the tools and equipment they were be broken down and reassembled for
and then kind of guess what’s happening
using. Rather than accept her fate, she new tasks once they’ve outlived their
in between. I was going after a technol-
hacked the machines until they finally original use.
did what she wanted. It got her thinking: Peek tries to make her tools as low- ogy that would fill that gap—what was
why couldn’t machines be more flexible? cost and accessible as possible: some happening to the cells throughout this
What if instead of changing your idea to use only cardboard for their frames, and transition.”
fit the tools, you could change the tools the designs are available to download. Platt has big ambitions for his tool.
to fit your idea? Thus began her quest to Her machines have been used by stu- He invented it to deal with a problem
create application-specific machines that dents, hackers, and even architects. that repeatedly frustrated him when he
could help anyone do almost anything. Peek’s goal is to give anyone with an was a graduate student at MIT. A group
Peek is now an assistant professor idea the means to translate it into phys- identified a gene that, when mutated
at the University of Washington, where ical reality. She notes that computers
and missing, appeared to play a role
she dedicates herself to this vision. She were originally designed to carry out
COURTESY PHOTO (PLATT)
in autism—though precisely when the

designs modular components—motors, specific tasks, but evolved to be more
gene affected the brain’s development
mechanical arms, and material cutters— general-purpose. She thinks machines
that can be assembled every which way that automate physical tasks should be remained a mystery.
and programmed with a little bit of code no different. “I ultimately really would like “If you want to identify a meaning-
to carry out tasks from the frivolous to to see automation as ... just another thing ful defect in a neuron you need to know
HERE
the scientific. When she teaches people that you can use for creative problem exactly when, where, and how to look,”
(PEEK);
CREDIT
to use her components, she delights in solving,” she says. —Karen Hao he says. “This was the biological problem
KERR
that motivated me to create the recording

GUTTER
SAM
tool.” —Adam Piore

Inventors 21
REBECCA SAIVE
University of Twente and ETC Solar
Age 33 | Country of birth: Germany

SIHONG WANG
She found a way to make solar University of Chicago
panels cheaper and more efficient.
Age 33 | Country of birth: China
The silver lines that crisscross the face

His stretchable microchips promise to make
of solar panels are essentially metal all sorts of new devices possible.
wires. They’re necessary to channel the
electric current flowing out of the cells,
but they reflect about 5% of the sunlight Microchips are usually machine. But that means called a “nanogenerator”—
that reaches them, creating the single etched into a substrate a set of new problems to rather than requiring exter-
biggest drain on their efficiency. of brittle silicon crystals. solve. How do you power nal batteries. Can these
Rebecca Saive, an assistant professor That means if you try to them? He’s already got then be placed inside the
bend or stretch them, their ways to harness energy body without triggering an
in applied physics at the University of
molecular structures break from the human body— immune response? That’s
Twente in the Netherlands, has invented
and performance drops using another invention next. —Bobbie Johnson
a novel type of “front contact” that
d ra m at i c a l l y. C i rc u i t s
addresses this problem, reducing the that aren’t as fragile have
wasted sunlight and improving the per- been around for a while,
formance of solar photovoltaics.
Her transparent contacts are made
but they’ve always had to
trade off performance or
VENKAT VISWANATHAN
from silver nanoparticles 3D-printed ease of manufacturing to
Carnegie Mellon University
onto the silicon layer of a solar cell, using achieve flexibility. Sihong
a technique she developed that produces Wang, however, has devel- Age 34 | Country of birth: India
an extremely thin and precise triangu- oped new manufacturing
lar shape. The steeply angled sidewalls techniques to build circuits His work on a new type of battery could make
reflect arriving light toward the absorb-
that can stretch and bend EVs much cheaper.
while performing just as
ing body of the cell like a mirror, boost-
well as an inorganic semi-
ing electricity output by at least 5% and conductor circuit. Venkat Viswanathan, an associate
lowering costs roughly the same amount. Building on his pre- professor at Carnegie Mellon, has
ETC Solar—a startup Saive cofounded vious work with Zhenan made major strides in developing
with headquarters in Pasadena, Bao at Stanford, one of the anodes made out of pure lithium,
California, and Rotterdam—produces a field’s pioneers, Wang has promising a new class of batteries that pack more
printing tool that enables manufacturers created a set of new pro- energy and deliver more power for a given amount
to integrate the technology into other- cesses that dramatically of weight. That could enable cheaper electric vehicles
wise standard photovoltaics. It’s already move things forward. Using and low-emissions aircraft.
a physical effect known as
selling the product, though the company Researchers have long recognized that lithium-metal
nanoconfinement to build
hasn’t announced customers yet. anodes could boost the performance of batteries over
layered polymer circuits at
Meanwhile, ETC and Saive’s aca- ones made of graphite. But they’re prone to devel-
the smallest possible scale,
demic team at the University of Twente he can now reliably build oping needle-like “dendrites” as lithium ions build
are using the front contacts and other high-performance circuits up. This can shorten the battery’s life and even spark
advances to develop even more effi- that can be stretched to fires. Viswanathan’s solution was developing a hybrid
cient solar cells that she says could twice their original length polymer-ceramic separator between the electrodes.
eventually lead to solar plants that pro- without losing any perfor- It applies enough pressure to prevent the dendrites
duce lower-cost electricity, and even to mance. from forming but still allows ions to flow through the
solar-powered cars. —James Temple These rubbery poly- battery, which produces the electric current.
mers, he says, open Viswanathan and colleagues secured more than
up whole new classes
$4 million from the Energy Department’s moonshot
of devices—malleable
ARPA-E program, and partnered with battery maker
enough to be molded to
24M Technologies to produce and test commercial-size
your shape, applied as a
skin patch, or even inserted lithium-metal cells.
inside the body, while able Viswanathan has also worked with Aurora Flight
HERE
PHOTOS
to do everything just as Sciences and Airbus A3 on battery designs for vertical

GUTTER CREDIT
well as a more traditional takeoff and landing aircraft, which can function as
COURTESY
air taxis or ambulances that zip across metropolitan

areas. —James Temple

Pioneers She’s helped

lead an effort
to bring
quantum
Their innovations lead the way to technology
biodegradable plastics, textiles that into the real
world.
keep you cool, and cars that “see.”
JENNIFER GLICK
IBM QUANTUM
with existing quantum algorithms or create In 2019, Glick and her colleagues
new ones for the purpose. tackled another big but more workaday
Quantum computing promises enor- problem with the banking giant Barclays.
If quantum computers work,
mous advances in processing power over The challenge was managing the quadril-
what can we use them for?
She’s working to figure that out. classical computing for certain problems lions of dollars processed each year in
that are intractably large or time-con- securities transaction settlements. These
The world’s biggest machine, the Large suming for classical computers—the kind occur, for instance, when a financial insti-
Hadron Collider, was built to help answer of problems Glick looks for. A quantum tution buys shares, bonds, or derivatives.
some of the most important questions in computer’s strength can be credited to Clearinghouses must run complex opti-
physics. To do that, the scientists behind the superposition and entanglement of mization algorithms on the transactions to
the particle collider have to be able to pro- quantum bits, or qubits, which offer an settle as many of them as possible within
cess and understand the massive amounts exponentially large computational space. technical and legal constraints.
of data from the machine. They want to be For example, 50 perfect qubits can repre- The results of the team’s research
able to tell whether certain particles are sent over a quadrillion states to explore. indicate that quantum technology could
produced in high-energy collisions taking Still, it’s a technology in its very early make this process more efficient, speed-
place at nearly the speed of light. days. In two years at IBM, Glick has helped ing up the time between trade and set-
The LHC can produce over a petabyte lead an effort to create partnerships that tlement. “When someone gives you an
of data per second from one billion parti- bring quantum technology into the real industry or business problem, there’s a
cle collisions, requiring about one million world. She spends a lot of her time hunt- lot of complications to start out with.
processor cores spread out around the ing for problems and then developing and It’s a very complex, gnarly problem,”
world to analyze and understand what demonstrating ways in which a quantum Glick says. “Part of it is breaking it
would otherwise be chaos. What does all computer could solve them faster than a down into simpler pieces to be able to
that data mean? classical one. identify where the bottlenecks are with
This is one of the most staggering “What we’re looking at for the Large respect to classical computing methods
problems facing Jennifer Glick, an IBM Hadron Collider is to use a quantum algo- that are being used today. And can any
researcher whose work is to find big prob- rithm to predict whether or not a certain of those bottlenecks be removed by an
lems that can benefit from quantum com- particle was produced,” she says. “Was that quantum approach?”
puting and then either try to solve them the particle I think was produced or not?” —Patrick Howell O’Neill
JA20_35-pioneers.indd 22 6/4/20 4:29 PM

Pioneers 23

LILI CAI
University of Illinois at Urbana-Champaign
in the brain. This happens to about half

She created energy-efficient textiles to break
our air-conditioning habit. of shunts within six years, so it’s a major
problem.
Earlier techniques for detecting shunt
Lili Cai has created polyethylene textile out how to fabricate failure all had various shortcomings.
nanomaterial-based tex- that can minimize heat her textiles in different Repeated CT scans, MRIs, or x-rays
tiles the thickness of a radiation loss but is still colors. Her goal is to subject patients to dangerous doses of
normal T-shirt that can breathable. Compared eventually produce one radiation, cost a lot, and—because they
keep you warm or cool with normal textiles, single adaptive textile measure the performance of shunts
you off. it keeps people about that keeps you warm if only indirectly—are not all that reli-
Cai’s work takes 7 °C warmer. Under it’s cold out, but cools able. Sometimes, invasive brain sur-
advantage of the direct sunlight, her you off in the heat. gery is done just to verify that a shunt
fact that human skin cooling fabric, a novel As climate change is working. And because checks were
strongly emits infrared nanocomposite mate- introduces shifts in being performed only a few times a year,
radiation in a specific rial, can cool the body weather and tempera- patients and their families had to live
range of wavelengths. by more than 10 °C. ture patterns globally, with constant uncertainty, wondering
By manipulating the Cai believes it’s people will use even if their shunts were working properly.
ways in which her fab- extremely important more energy to regulate In any case, because the flow of fluid
rics block or transmit to figure out how to building temperatures. from the brain is naturally intermit-
radiation in this band, make such textiles look If she can figure out how tent, spot checks don’t necessarily catch
she has produced mul- as much as possible to cheaply make her tex- problems.
tiple textiles that can like normal clothing. tiles at scale, they will Krishnan’s sensor offers a noninvasive
have different effects Previous radiative cool- provide an alternative way to monitor the flow in shunts: it can
on temperature. ing materials could only that could help cut those be placed over the skin on the neck, near
To heat the body, be produced in white, heating and cooling the valve. It measures the temperature at
Cai created a metallized but in 2019 Cai figured bills. —Abby Ohlheiser several distinct spots, inferring from the
temperature distribution at those spots
whether or not liquid is flowing. Unlike
an earlier generation of noninvasive sen-
sors, which made fewer temperature
SIDDHARTH KRISHNAN
Between one and two in every 1,000 measurements and required the use of
babies born in the United States have an ice pack, his device can continuously
hydrocephalus, a condition in which cere- measure the flow, reporting results via
MIT brospinal fluid builds up in the brain. It Bluetooth.
can also occur later in life, including after So far, field trials on seven patients
Age 29 | Country of birth: India
traumatic brain injury. Over a million peo- reported in a paper earlier this year in
ple in the United States have hydroceph- the journal NPJ Digital Medicine show
A tiny, powerful sensor for making
alus, and nearly all of them have a shunt that his sensor gives “robust, high-quality
disease diagnosis cheaper, faster,
and easier. installed that drains fluid from their brain data” for hours at a time.
into their chest or abdomen. The condition Krishnan hopes that his sensor will
COURTESY PHOTO (CAI); EILEEN MOLONY (KRISHNAN)
Siddharth Krishnan, a materials scientist can be fatal if untreated, but if it’s dealt with have applications beyond hydrocephalus,
at MIT, developed a tiny sensor that could promptly a full recovery is often possible. possibly monitoring other diseases like
save people from a devastating and often If shunts fail, because they get diabetes, where tiny changes beneath the
deadly brain condition. clogged, then fluid will again build up skin can have huge effects. —Tanya Basu
His noninvasive sensor can radically improve the treatment

for hydrocephalus, which can be fatal if left untreated.

Pioneers 25
AVINASH MANJULA
BASAVANNA “If something
Wyss Institute, Harvard University
were to
Age 33 | Country of birth: India
happen, and
His biodegradable plastic protects desalination
against extreme chemicals, but
heals itself using water.
plants
weren’t able
Of the estimated 9.1 billion tons of plastic to operate ...
ever produced, only 9% has been recy-
cled. Almost 80% ends up as waste that
there really
adds to growing landfills or pollutes the is no backup
natural environment, where it takes a plan.”
thousand years to degrade. Such mate-
rials can also end up in the human body
as microplastics, slowly accumulating
with devastating effects on health. One
key to solving these problems could be
bioplastics—plastic alternatives pro-
duced through bioengineered organ-
isms. These can degrade naturally and
much more quickly.
The idea of bioplas- GHENA ALHANAEE
tics isn’t exactly new, University of Southern California
but it’s been difficult to
make them in the sorts Age 30 | Country of birth: United Arab Emirates
of quantities and with

the properties that would Heavy dependence on infrastructure like oil rigs, nuclear reactors,
be useful for industry. Avinash Manjula
and desalination plants can be catastrophic in a crisis. Her data-
driven framework could help nations prepare.
Basavanna, a postdoc at the Wyss Institute
for Biologically Inspired Engineering at Early on in her days as a doctoral stu- better mitigate the risks of an oil spill
Harvard University, thinks he can do bet- dent at the University of Southern Cali- or nuclear accident. Since the Gulf’s
ter. He and his colleagues have developed fornia, Ghena Alhanaee stumbled upon nuclear industry is nascent, and its oil
a new type of plastic based on living mate- a disturbing set of facts. The countries and gas sector keeps its data private,
of the Persian Gulf, including her native she’s relying on information from the
rials that he calls AquaPlastic and which
United Arab Emirates, were far more vul- US: her statistical model draws on data
can be produced at a commercial scale,
nerable to disaster than she’d realized. from more than 4,000 reported safety
exhibits the tough qualities of many petro-
Not only was the Gulf itself one of the incidents in the US nuclear and offshore
leum-based plastics, and can degrade in world’s largest oil and gas production oil industries over the past decade. The
water in as little as two months. zones, with more than 800 offshore plat- trick, she says, is to better understand
The material itself is resistant to strong forms and thousands of tankers passing which combinations of small incidents,
acids and bases. It can be applied as a through its shallow waters every year, but under which scenarios, are most likely
coating using nothing but water, which the UAE was also building the Arab Pen- to snowball into something major.
makes the plastic turn adhesive—the insula’s first nuclear power plant. Mean- Alhanaee’s framework seeks to do
first plastic of its kind to boast this fea- while, several Gulf countries relied almost just that. She plans to apply her find-
(BAVANASSA); SAM KERR (ALHANAEE)
ture. If it gets scratched, the coating can exclusively on desalinated Gulf water for ings to a particularly vulnerable spot in
drinking, with emergency supplies for just the Gulf—in the vicinity of the Barakah
also be “healed” using water. And most
two or three days. “If something were to nuclear power plant, which is nearing
important, “it’s flushable,” says Manjula
happen, and desalination plants weren’t completion, and large-scale oil and
Basavanna. “You don’t have to worry about
able to operate, right now there really is desalination installations. Ultimately,
it adding to our plastic and microplastic no backup plan,” Alhanaee says. she hopes her research will help the
problem.” He and his partners are now in Ever since, she has devoted her region’s governments develop more
the beginning stages of forming a startup energy to tackling the Gulf’s disaster robust, and better coordinated, disaster
HERE
around AquaPlastic. If manufactured at

PHOTO
preparedness gap. She’s developing a mitigation strategies.

GUTTER CREDIT
scale, the cheap, biodegradable material data-driven framework to help the region —Jonathan W. Rosen
COURTESY
could compete with conventional plastic

coatings. —Neel V. Patel
GREGORY EKCHIAN
MIT
Using Age 32 | Country of birth: US
Karpathy’s He invented a way to make

advances, radiation therapy for cancer safer
Tesla is and more effective.
taking a The amount of radiation it takes to kill
different a tumor depends on the level of oxygen

in the tumor cells. This can vary greatly,
path from but oncologists don’t currently adjust
most other radiation doses to account for it. Gregory
Ekchian, cofounder of Stratagen Bio, has
automakers. developed a sensor for reading tumor oxy-
gen levels to personalize cancer treatment.
Ekchian recognized a glaring need
for a new sensing tool after discussions
with clinicians at Brigham and Women’s
Hospital in Boston. He developed a pro-
totype for a cancer treatment technique
called high-dose-rate brachytherapy. In
this form of treatment, doctors punc-
ture the tumor with a series of hollow
ANDREJ KARPATHY catheter tubes and then drop radioac-

tive seeds through the
Tesla tubes to suffuse the
tumor with radiation,
Age 33 | Country of birth: Slovakia removing them once the
desired dose has been
He’s employing neural networks to allow automated cars to “see.” delivered.
For his prototype, Ekchian added a
strip of a recently invented oxygen-sen-
Getting computers to see—to actually In 2017, Karpathy joined Tesla, where sitive polymer to the tips of a modified
see—has been an ambition of countless he oversees neural networks for the version of the catheters. During routine
computer scientists for decades. Few cars’ Autopilot feature. That includes MRI scans, protons in the polymer are
have come closer than Andrej Karpathy, collision detection, self-driving capabili- excited; these protons return to equilib-
whose approach to deep neural net- ties, and summoning (having a car drive
rium far faster in catheters surrounded
works allows machines to make sense autonomously from where it is parked).
by high levels of oxygen than low levels.
of what is happening in images. Using Karpathy’s advances, Tesla
The speed at which they return to equi-
As a graduate student at Stanford, is taking a different path from most
Karpathy extended techniques for build- other automakers. Typically, self-driving librium can therefore be used to map
ing what are known as convolutional vehicles scan their surroundings with out how oxygen levels vary in different
neural networks (CNNs)—systems that expensive laser range finders, build a parts of the tumor, allowing oncologists
broadly mimic the neuron structure in the virtual map, and then use AI to make to pinpoint where radiation doses should
visual cortex. (In 2015 he also designed decisions about what to do. Tesla’s go and tailor their length and intensity to
and was the primary instructor for the approach uses traditional cameras. Not be most effective.
first deep-learning class at Stanford.) only can Karpathy’s method let the car “If we weren’t worried about healthy
By combining CNNs with other spot objects in the road as a human tissue, we would just boost the dose to
HERE COURTESY PHOTO (EKCHIAN)
deep-learning approaches, he created a driver would, but it can take in the entire
the entire tumor,” he says, but excess radi-
system that was not just better at recog- scene (cars, people, intersections, stop
ation can harm the patient. That means
nizing individual items in images (say, a signs, and more) and—if it works as
“it’s really important to figure out where
dog or a person), but capable of seeing intended—instantly infer what’s taking
an entire scene full of objects—multiple place. Doing so requires nearly 50 neu- those high doses need to go.”
dogs and people interacting with each ral networks to constantly process data Ekchian is preparing to publish the
results of a clinical trial involving seven
(KARPATHY);
other—and effectively building a story of coming in as the more than a million

what was happening in it and what might cars in the fleet look and learn. patients with cervical cancer, the first in
CREDIT
happen next. —Bobbie Johnson humans. He ultimately hopes to employ

KERR
his oxygen-sensing applications for a wide

GUTTER
SAM
range of clinical needs. —Adam Piore

Pioneers 27
broad-acting antiviral drugs. “The idea

is that viruses depend on specific cel-
lular pathways which could themselves
become drug targets,” says Puschnik.
Usually, the German-born
researcher says, drug makers look to
take out pathogens with chemicals
designed to bind to and disable the
molecular components of the virus
itself. This “one drug, one bug” solution
can work powerfully (think HIV drugs).
The problem is that each drug has to
be specially designed.
An alternative, called host-directed
therapeutics, is in its early days. But
Puschnik has helped speed it up using
the gene-editing tool CRISPR. In a
mass screening approach, he uses
CRISPR to pepper millions of human
cells growing in flasks with a hundred
thousand different genetic mutations.
If any of those cells survive infection
with, say, yellow fever, it means he’s
inactivated a molecular pathway the
germ needs to reproduce.
Puschnik has already helped find an
enzyme that mosquito-borne flavivi-
ruses like dengue, Zika, and West Nile
need to reproduce, as well as a drug
to block it. Since all flaviviruses work
similarly, he hopes the drug could be a
“universal treatment” for them.
During California’s 2020 lockdown,
ANDREAS PUSCHNIK
the biologist remained at work at the
Chan Zuckerberg Biohub, a new insti-
tute that picked him as its first sci-
Chan Zuckerberg Biohub entific fellow. “It is still busy days for
Zika, Ebola, SARS, dengue fever, and virologists,” says Puschnik, who now
covid-19. These diseases have fearsome plans to turn his attention to the coro-
personalities, yet the viruses that cause navirus that causes covid-19. Perhaps,
Seeking a universal treatment
them are not really alive. To reproduce, he thinks, a drug that changes cells
for viral diseases, he might leave us
much better prepared for the next viruses need to hijack a cell and use its so they are less hospitable to corona-
GUTTER CREDIT HERE
pandemic. components to produce more viruses. viruses could be ready for the next
To Andreas Puschnik, understand- pandemic: “You might be able to treat
ing which of our biomolecules viruses viruses you don’t even know about yet.”
depend on could lead to new types of —Antonio Regalado

little of the country she left other than the

reason for leaving: her family wanted to
escape its instability and give her and her
siblings a better life. The transition proved
tough. For the first two years, Raji’s father
Visionaries
continued to work in Nigeria, flying back
and forth between two continents. Raji
attended seven different schools during
their first five years in Canada.
Eventually, the family moved to Ottawa
and things began to stabilize. By the time
Their innovations are leading to she applied to college, she was sure she was
most interested in pre-med studies. “I think
breakthroughs in AI, quantum if you’re a girl and you’re good at science,
people tell you to be a doctor,” she says.
computing, and medical implants. She was accepted into McGill University
as a neuroscience major. Then, on a whim,
and with her father’s encouragement, she
visited the University of Toronto and met
a professor who persuaded her to study
INIOLUWA DEBORAH RAJI

engineering. “He was like, ‘If you want to
use physics and you want to use math to
build things that actually create impact,
you get to do that in this program,’” she
AI Now Institute Though Raji told Clarifai about the remembers. “I just fell for that pitch and
problem, the company continued using overnight changed my mind.”
Age 24 | Country of birth: Nigeria
the model. “It was very difficult at that It was at university that Raji took her
time to really get people to do anything first coding class and quickly got sucked
Her research on racial bias in data
used to train facial recognition about it,” she recalls. “The sentiment was into the world of hackathons. She loved
systems is forcing companies to ‘It’s so hard to get any data. How can we how quickly she could turn her ideas into
change their ways. think about diversity in data?’” software that could help solve problems
The incident pushed Raji to investi- or change systems. By her third year, she
The spark that sent Inioluwa Deborah gate further, looking at mainstream data was itching to join a software startup
Raji down a path of artificial-intelligence sets for training computer vision. Again and experience this in the real world.
research came from a firsthand realiza- and again, she found jarring demographic And so she found herself, a few months
tion that she remembers as “horrible.” imbalances. Many data sets of faces lacked into her internship at Clarifai, search-
Raji was interning at the machine- dark-skinned ones, for example, leading ing for a way to fix the problem she had
learning startup Clarifai after her third to face recognition systems that couldn’t discovered. Having tried and failed to
year of college, working on a computer accurately differentiate between such get support internally, she reached out
vision model that would help clients flag faces. Police departments and law enforce- to the only other researcher she knew
inappropriate images as “not safe for ment agencies were then using these same of who was working on fighting bias in
work.” The trouble was, it flagged photos systems in the belief that they could help computer vision.
of people of color at a much higher rate identify suspects. In 2016, MIT researcher Joy Buolamwini
than those of white people. The imbal- “That was the first thing that really (one of MIT Technology Review’s 35
ance, she discovered, was a consequence shocked me about the industry. There are Innovators Under 35 in 2018) gave a TEDx
of the training data: the model was learn- a lot of machine-learning models currently talk about how commercial face recog-
ing to recognize NSFW imagery from being deployed and affecting millions and nition systems failed to detect her face
porn and safe imagery from stock pho- millions of people,” she says, “and there unless she donned a white mask. To Raji,
tos—but porn, it turns out, is much more was no sense of accountability.” Buolamwini was the perfect role model:
diverse. That diversity was causing the Born in Port Harcourt, Nigeria, Raji a black female researcher like herself who
model to automatically associate dark moved to Mississauga, Ontario, when she had successfully articulated the same prob-
skin with salacious content. was four years old. She remembers very lem she had identified. She pulled together
JA20_35-visionaries.indd 28 6/4/20 5:11 PM

Visionaries 29
all her code and the results of her analyses processes, but the testing data they used how well a system recognizes people from
and sent Buolamwini an unsolicited email. was as demographically imbalanced as the different demographic groups.
The two quickly struck up a collaboration. training data the systems learned from. Raji joined in the technical work,
At the time, Buolamwini was already As a result, the systems could perform compiling the new data set and helping
working on a project for her master’s the- with over 95% accuracy during the audit Buolamwini run the audits. The results
sis, called Gender Shades. The idea was but have only 60% accuracy for minority were shocking: among the companies
simple yet radical: to create a data set that groups once deployed in the real world. they tested—Microsoft, IBM, and Megvii
could be used to evaluate commercial By contrast, Buolamwini’s data set would (the company best known for making the
face recognition systems for gender and have images of faces with an even distri- software Face++)—the worst identified
racial bias. It wasn’t that companies selling bution of skin color and gender, making the gender of dark-skinned women 34.4%
these systems didn’t have internal auditing it a more comprehensive way to evaluate less accurately than that of light-skinned

“There are a lot of machine-learning

EIMEAR DOLAN
models currently being deployed ... and National University of Ireland Galway
there was no sense of accountability.” Age 32 | Country of birth: Ireland
Medical implants are often

thwarted as the body grows tissue
to defend itself. She may have found
a drug-free fix for the problem.
men. The other two didn’t do much Clarifai to make sure it would be easy to
better. The findings made a headline adhere to. Google rolled out the frame-
in the New York Times and forced the work in 2019 and built it into Google When Eimear Dolan first worked to
companies to do something about the Cloud for its clients to use. A number develop implantable medical devices
bias in their systems. of other companies, including OpenAI to treat type 1 diabetes, she and her col-
Gender Shades showed Raji how and natural-language processing firm leagues had to overcome a common
auditing could be a powerful tool for Hugging Face, have since adopted roadblock. Their problem was one that’s
getting companies to change. So in the similar practices. long dogged makers of devices like pace-
summer of 2018, she left Clarifai to pur- Raji also co-led her own project at makers, insulin delivery systems, and
sue a new project with Buolamwini at Google to introduce internal auditing breast implants: when the body senses
the MIT Media Lab, which would make practices as a complement to the exter- an implanted foreign object, it constructs
its own headlines in January 2019. This nal auditing work she did at the Media a protective wall of fibrous tissue. This
time Raji led the research. Through Lab. The idea: to create checks at each reaction, known as the foreign body
interviews at the three companies stage of an AI product’s development response, is one of the main reasons
they’d audited, she saw how Gender so problems can be caught and dealt medical implants fail.
Shades had led them to change the ways with before it is put out into the world. Today, as a biomedical engineer at the
they trained their systems in order to The framework also included advice on National University of Ireland Galway,
account for a greater diversity of faces. how to get the support of senior man- Dolan thinks she’s found a way to coun-
She also reran the audits and tested two agement, so a product would indeed teract the foreign body response. Her
more companies: Amazon and Kairos. be held back from launching if it didn’t weapon is a small robotic device known
She found that whereas the latter two pass the audits. as a dynamic soft reservoir. Developed
had egregious variations in accuracy With all her projects, Raji is driven through a collabora-
between demographic groups, the orig- by the desire to make AI ethics easier to tion between Dolan’s
inal three had dramatically improved. practice—“to take the kind of high-level lab at NIU Galway and
The findings made a foundational ethical ideals that we like to talk about researchers at MIT,
contribution to AI research. Later as a community and try to translate that the device is made of a
that year, the US National Institute of into concrete actions, resources, and soft material that can be
Standards and Technology also updated frameworks,” she says. made to oscillate, creating enough fluid
its annual audit of face recognition algo- It hasn’t always been easy. At Google, flow to alter the environment around
rithms to include a test for racial bias. she saw how much time and effort it the implant and keep protective tissue
Raji has since worked on several took to change the way things were from forming.
other projects that have helped set stan- done. She worries that the financial Past researchers have sought to use
dards for algorithmic accountability. cost of eliminating a problem like AI drugs or modify the surface chemistry
After her time at the Media Lab, she bias deters companies from doing it. of an implant. Dolan’s innovation, which
joined Google as a research mentee to It’s one reason she has moved back she and her colleagues have successfully
help the company make its AI devel- out of industry to continue her work tested in rats, marks the first time anyone
opment process more transparent. at the nonprofit research institute AI has tackled the problem mechanically.
Whereas traditional software engineers Now. External auditing, she believes, “The beauty about it is it’s a drug-free
have well-established practices for doc- can still hold companies accountable approach,” Dolan says.
umenting the decisions they make while in ways that internal auditing can’t. Her team is redesigning the dynamic
building a product, machine-learning But Raji remains hopeful. She sees soft reservoir as part of an effort to
engineers at the time did not. This made that AI researchers are more eager construct a “bioartificial pancreas,” an
it easier for them to introduce errors or than ever before to be more ethical implantable reservoir of cells that pro-
bias along the way, and harder to check and more responsible in their work. duce insulin for people with type 1 diabe-
such mistakes retroactively. “This is such impactful technology,” tes. Early attempts at such devices have
Along with a team led by senior she says. “I just really want us to be been particularly liable to be rejected
research scientist Margaret Mitchell, more thoughtful as a field as to how by the body and fail. Dolan believes her
LILLIE PAQUETTE
Raji developed a documentation frame- we build these things, because it does team can change that—and ultimately
work for machine-learning teams to matter and it does affect people.” improve the success of other implantable
use, drawing upon her experience at —Karen Hao devices. —Jonathan W. Rosen

Visionaries 31
MIGUEL MODESTINO
NYU
Age 34 | Country of birth: Venezuela
He is reducing the chemical industry’s

carbon footprint by using AI to optimize
reactions with electricity instead of heat.
ZLATKO MINEV
Miguel Modestino has system teaches itself how which typically involves
cleared a major hurdle to optimize the reactions burning fossil fuels. And
in electrifying the chem- for making various chem- since electricity can come
ical industry, which pro- icals by zapping them from renewable sources IBM Quantum Research, TJ Watson
duces compounds used with pulses of electricity like wind or solar, electrify-
in everything from plastics instead of the conventional ing chemical plants could Age 30 | Country of birth: Bulgaria
to fertilizer. His AI-based approach of heating them, greatly reduce emissions.
In an early lab proj- His discovery could reduce errors
ect, Modestino’s team in quantum computing.
achieved more than a
ROSE FAGHIH 30% boost in the pro-

duction rate of adipo-
Zlatko Minev overturned a mainstay of
quantum physics that had troubled Niels
University of Houston and MIT
nitrile (which is used in Bohr and Albert Einstein alike. For most
making nylon, among of the 20th century, it was assumed that
Age 34 | Country of birth: US numerous other industrial atoms change from one energy level to
processes)—a greater another in abrupt, unpredictable, discrete
Her sensor-laden wristwatch would monitor improvement than any
quantum jumps. Minev proved otherwise.
your brain states. other method has shown
EDUARDO WAITE (MODESTINO); JEFF LAUTENBERGER, CULLEN COLLEGE OF ENGINEERING, UNIVERSITY OF HOUSTON(FAGHIH); ROBERT JONES (MINEV)
“Quantum physics is not quite as

in the last 50 years.
If Rose Faghih’s project pans out, a unpredictable and discrete as we previ-
The key was using
seemingly simple smart watch could complex pulses of electri- ously thought,” he says.
determine what’s happening deep cal current at constantly His experiment showed that when an
inside your brain. varying rates to optimize atom is bombarded with energy in the form
Faghih has developed an algorithm yields. Figuring out what of light, it moves from one energy level
to analyze otherwise imperceptible patterns of pulses to use to the next in a continuous, smooth way,
changes in sweat activity—a key indicator of stress required machine learn- not an instantaneous jump. What’s more,
and stimulation. Using two small electrodes attached ing. Modestino ran a few Minev was able to detect the change in an
to the back of a smart watch, she can monitor changes experiments making adi- atom’s energy level quickly enough to con-
ponitrile under different
in skin conductance caused by sweat. Signal-processing trol it so he could stop the jump midflight
electrical conditions and
algorithms then allow Faghih to correlate those changes and reverse it before it was completed.
then let his AI analyze the
with specific events, such as a PTSD-related flashback “In the short term,” he says, “with
data to figure out how to
or even just wandering attention, in order to pinpoint make the compound with the monitoring that I developed for this
the person’s brain state. less energy, better yields, project, we can actually have a window
Typically, this kind of real-time data is available only and less waste. of predictability.”
by way of expensive scalp-based electrode systems like Modestino and two Minev’s work could have major impli-
EEG or functional MRI. Faghih’s “Mindwatch” would former students recently cations for quantum computing. Such
in theory be cheap and portable enough to let people founded Sunthetics systems are riddled with errors that occur
monitor their brain states anywhere. to apply the AI sys- when subatomic particles jump between
Faghih hopes it will help people manage their own tem to other chemical energy levels, like the atoms in Minev’s
processes, like those
changing moods and mental states: a wearable with experiment. The ability to detect and
involved in generating
her technology could suggest that an agitated driver reverse such jumps before they finish
hydrogen fuel and making
try some deep breathing or prompt a lonely shut-in should dramatically boost the power of
polymers. The company is
to turn on mood-enhancing music. For people with also working to scale up quantum computers, allowing them to
mental illness or chronic conditions like diabetes, it the adiponitrile process better crack encryption, model chemical
could potentially even trigger an automated deep-brain for a full pilot reactor and reactions, and forecast weather.
stimulation device or an insulin pump. —Kathryn Miles to extend the approach to —Russ Juskalian
other processes.
—Russ Juskalian

DONGJIN SEO
Neuralink
Her goal is Age 31 | Co

Country of birth: US
to use her He is designing computer chips to

knowledge seamlessly connect human brains
about and machines.
potential When I first met the electrical engineer
attacks to Dongjin “DJ” Seo six years ago, he told

me he had always wanted to be “a sci-
make AI entist with strong intuitions about how
more robust. to improve the world through engineer-
ing.” At the time, he was working in a
crowded corner of a lab at the University
of California, Berkeley, on a concept
called neural dust—ultra-small elec-
tronic sensors that could be sprinkled
in an animal’s brain and controlled with
acoustic waves.
The goal of that project was new types
of brain-machine interfaces that could
read the firing of neurons inside the
cortex and even send information back
BO LI in. That kind of technology might open

up ways to read and write information
University of Illinois at Urbana-Champaign from and to the brain.
Then, in 2016, not too long after I
Age 32 | Country of birth: US spoke with Seo, Elon
Musk tapped him to
By devising new ways to fool AI, she is making it safer. join a new company,
Neuralink, which was
A few years ago, Bo Li and her col- and texture, that again are impercepti- ready to spend millions
leagues placed small black-and-white ble to humans but can make the objects on engineering a seam-
stickers on a stop sign in a graffiti-like invisible to image recognition algo- less interface between human brains and
pattern that looked random to human rithms. Her goal is to use this knowledge computers. “The vision that Elon out-
eyes and did not obscure the sign’s about potential attacks to make AI more lined—well, it was hard to say no,” Seo
clear lettering. Yet the arrangement robust. She pits AI systems against
says. “It was everything I had imagined.”
was deliberately designed so that if an each other, using one neural network
Instead of neural dust, the startup is
autonomous vehicle approached, the to identify and exploit vulnerabilities in
betting on a robot that plunges ultra-
neural networks powering its vision sys- another. This process can expose flaws
tem would misread the stop sign as one in the training or structure of the target thin electrodes into animal brains. Seo is
posting a speed limit of 45 mph. network. Li then develops strategies to head of a team of about a dozen people
Such “adversarial attacks”—manip- patch these flaws and defend against designing low-power wireless comput-
ulation of input data that looks innoc- future attacks. ers that fit into a small burr hole that’s
uous to a person but fools neural Adversarial attacks can fool other cut into the skull. He says his primary
networks—had been tried before, but types of neural networks too, not just contribution is designing the necessary
earlier examples had been mostly dig- image recognition algorithms. Impercep- circuit boards and chips. “We need these
ital. For instance, a few pixels might be tible tweaks to audio can make a voice chips to collect a signal that may look
altered in an image, a change invisible to assistant misinterpret what it hears, for
like noise, process it, and do all that
the naked eye. Li was one of the first to example. Some of Li’s techniques are
without cooking your brain.”
show that such attacks were possible in already being used in commercial appli-
PHOTO (SEO)
After tests on animals, the company

the physical world. They can be harder cations. IBM uses them to protect its
for an AI to detect because the methods Watson AI, and Amazon to protect Alexa. hopes to try the brain connection on
developed to spot manipulated digital And a handful of autonomous-vehicle someone with paralysis or a serious
(LI); COURTESY
images don’t work on physical objects. companies apply them to improve the illness. Eventually, “augmentation” of
HERE
Li also devised subtle changes in the robustness of their machine-learning healthy people “is an obvious result,”
CREDIT
features of physical objects, like shape models. —Will Douglas Heaven Seo says: “It’s being able to enhance our
KERR
ability to interact with the world.”

GUTTER
SAM
—The editors

Visionaries 33
LEILANI BATTLE
University of Maryland

MORGAN BELLER
Her program sifts through data Novi
faster so scientists can focus more
on science. Age 27 | Country of birth: US
When Leilani Battle was working on her

She was a key player behind the idea of a
PhD, she helped develop ForeCache, a Facebook cryptocurrency.
tool designed to help researchers browse
large arrays of data—for instance, scan-
ning high-resolution satellite images to In the summer of 2017, Mor- Libra hasn’t even rolled recently announced plans
look for areas covered with snow. The gan Beller approached her out yet, but it’s already to scale back Libra and first
goal is to reduce latency, so that a user supervisor on Facebook’s prompted several coun- issue a coin backed by a
can pan and zoom across the data set corporate development tries, including China, to local currency, but even with
team with a proposal: what accelerate the develop- these modifications, Libra
without perceptible delay. A common
if she began spending the ment of their own national has already been disruptive.
way to do this is to predict which parts
bulk of her job researching digital currencies. Facebook —Kathryn Miles
of the data a user is likely to need and
how the social-media giant
then “prefetch” them. But how to pre- could enter the digital cur-
dict what to prefetch? That depends on rency market?
understanding the user’s behavior.
Battle and her colleagues developed
Beller was so new at
Facebook that she was
ADRIANA SCHULZ
a more efficient prediction system. It still completing her orien-
University of Washington
attempts to discern first which “analysis tation, but she’d cut her
phase” a user is in, and then what tiles of teeth at a venture capital Age 34 | Country of birth: Brazil
data might be wanted next. They dubbed firm, where she’d worked on
the three phrases “foraging,” “sensemak- early cryptocurrency invest- Her tools let anyone design products without
ing,” and “navigation.” They suppose that
ments. She could see that having to understand materials
users in the “foraging” phase are browsing
a seismic shift in the global science or engineering.
financial community was
at a coarse level, in order to come up with coming. Adriana Schulz’s computer-based
new ideas. “Sensemaking” is a closer exam- When she realized that design tools let average users and
ination meant to test those ideas, and “nav- no one at Facebook was engineers alike use graphical drag-
igation” is a transition between the two. working on blockchain, she and-drop interfaces to create func-
This system allowed them, they said, volunteered and quickly tional, complex objects as diverse as robots and
to predict which tiles users wanted about became the company’s birdhouses without having to understand their
25% better than existing prefetching sys- digital currency evan- underlying mechanics, geometries, or materials.
tems they benchmarked against, almost gelist, shepherding the “What excites me is that we’re about to enter the
development of both its
halving the latency. next phase in manufacturing—a new manufacturing
open-source blockchain
Battle has devoted her career to revolution,” says Schulz.
infrastructure, Libra, and
designing systems and interfaces that One of her creations is Interactive Robogami,
its currency application and
help researchers sifting through data do digital wallet, Novi. Today a tool she built to let anyone design rudimentary
their work better and faster. She hopes she serves as head of strat- robots. A user designs the shape and trajectory of
to make exploration tools more interac- egy for the latter, where she a ground-based robot on the screen. Schulz’s sys-
tive and visual so they’ll be less daunt- works with a team of digital tem automatically translates the raw design into
ing. Perhaps this will allow scientists to currency developers. a schematic that can be built from standard or
(BATTLE, BELLER); DAVID CURTIS (SCHULZ)
spot data quirks that would otherwise go Fa c e b o o k a n d i t s 3D-printed parts.

unnoticed. —Tanya Basu founder, Mark Zuckerberg, Another of the tools she and her collaborators
endured sharp criticism built lets users design drones to meet their chosen
after announcing the plans
requirements for payload, battery life, and cost. The
for Libra. Beller wasn’t sur-
algorithms in her system incorporate materials sci-
prised. “We’re trying to
ence and control systems, and they automatically
change the system, and
there are a lot of people output a fabrication plan and control software.
who are incentivized for the Schulz is now helping start the University of
HERE
PHOTOS
global financial system not Washington Center for Digital Fabrication, which
GUTTER CREDIT
to change,” she says. she will co-direct. She will work with local tech-
COURTESY
nology and manufacturing companies to move her

tools out of the lab. —Russ Juskalian

papers and data sets published on a topic

and identifying insights that could lead to
breakthroughs. Though machine learning
isn’t her specialty, she brought together a
team of AI researchers, along with experts
Humanitarians
from other fields like computational biol-
ogy, drug development, and neurosci-
ence. She raised money from various
investors, including Jeff Dean (the head
of AI at Google) and the Michael J. Fox
Foundation. Thus, in 2016, OccamzRazor
They’re using technology to cure was born.
The company is tackling the problem
diseases and make water, housing, in two major steps. First, it has developed
programs that read and understand pub-
and prosthetics available to all. lished materials on Parkinson’s. Next, it

is using AI to integrate genomics, pro-
teomics, and clinical data sets. The goal
is to predict new pathways and genes
important to Parkinson’s that can then
KATHARINA VOLZ
be tested in the laboratory.
The result is what OccamzRazor calls
the “Parkinsome”—a knowledge map
OccamzRazor of Parkinson’s that reveals how the dis-
generally didn’t know much about and ease is caused and progresses, points
couldn’t engage with other aspects. to signs and symptoms that can help
These academic silos made it hard for make an early diagnosis, and identi-
A loved one’s diagnosis led her
new insights to be properly shared and fies potential therapeutic targets. After
to employ machine learning in the
search for a Parkinson’s cure. explored, impeding our continued under- OccamzRazor validates its findings, it
standing of how Parkinson’s progresses. partners with biotech and pharma com-
In 2016, Katharina Volz received news “Even if you’re the smartest researcher panies to develop drugs.
that someone close to her had Parkinson’s. in the world, you can’t put all of this The goal is to take this approach
At the time Volz had just finished her information together and make the con- beyond just Parkinson’s. Volz and her
PhD at Stanford and was locked into a nections you need to truly understand team have plans to scale up the platform
well-earned career in academic research, how the disease operates,” says Volz. “As to build comprehensive knowledge maps
working on stem cells. But the news humans, our ability to draw these numer- for other complex diseases related to the
changed all that. ous connections is limited.” aging of the brain. “Diseases inform each
“I just knew I could actually make a That’s where machine learning comes other,” says Volz. “Studying Parkinson’s is
difference,” she says. “Sometimes you in. Volz realized AI could do a better job one of the best ways to study brain aging
feel helpless. But actually I felt deeply than a human at reading all the different in general.” —Neel V. Patel
responsible for finding a way to get cura-
tive treatments for this disease, because I
knew I could do something about it.” Volz
now leads a company, OccamzRazor, that
has successfully married machine learning “I felt deeply responsible for finding a way to
with biomedical research and is pushing
the search for a Parkinson’s cure.
get curative treatments for this disease,
Volz noticed a problem when it comes because I knew I could do something about it.”
to researching Parkinson’s, and it’s one
that arguably plagues science at large.
Experts studying the disease were spe-
cializing in particular aspects of it and
JA20_35-humanitarians.indd 34 6/5/20 1:02 PM

Humanitarians 35

ALEX Le ROUX
Icon
“Globally, Age 27 | Country of birth: Canada
we are truly A massive 3D-printing project in

tapping out Mexico could point the way to the
our water future of affordable housing.
resources.”
Alex Le Roux thinks 3D printing can
open new possibilities for architectural
design and cut the cost of building hous-
ing around the world.
As cofounder of Icon, a startup based in
Austin, Texas, Le Roux is the mastermind
behind the Vulcan, an industrial-scale 3D
printer that can construct the wall system
of an entire house in just 24 hours of print
time. According to the United Nations,
some 1.6 billion people lack adequate
shelter, and a third of the world’s urban
population lives in informal settlements or
slums. Part of the reason, Le Roux says, is
that traditional building methods lead to
DAVID WARSINGER wasted materials and excess labor costs,

driving up housing prices beyond the
Purdue University reach of many poor families.
The Vulcan is designed to change that
Age 32 | Country of birth: US by introducing automation to the process.
The 12-foot-tall robotic device works by
His system could alleviate the drawbacks of existing desalination plants. extruding inch-thick layers of a special con-
David Warsinger thinks he’s found an uses a lot of energy and is costly. It crete mix fed in from a separate machine,
innovation that could help combat one also leaves behind a large part of the much like a giant tube of toothpaste. Icon
of the 21st century’s great environmen- water as brine—an especially big prob- programs its home designs ahead of time
tal challenges: water shortages around lem for inland plants, where source to make the operator’s job as simple as
the globe. water is scarcer. possible. “Once these two machines are
His fix is an improved form of Warsinger’s system, which he devel- set up on a job site, you download an app
reverse osmosis—the most common oped with Emily Tow while they were
and you’re off to the races,” Le Roux says.
method of desalination. Today, an esti- both at MIT, is known as batch reverse
In March 2018, Icon built the US’s first
mated 5% of the world’s population osmosis, and it is designed to make the
officially permitted 3D-printed house. It
relies on desalinated water, drawn from process more efficient. The technique
the ocean or brackish inland sources, allows desalination to occur in batches, has now built 16 houses in Austin and in
to meet at least some daily needs. This with salinity and pressure varying over Mexico, where it’s constructing the world’s
figure will continue to rise as aqui- time. Whereas traditional reverse osmo- first 3D-printed community, designed to
fers are further squeezed by pollution, sis systems apply constant pressure, accommodate 50 low-income families.
SAM KERR (WARSINGER); NATALIE CASS, SCORE HEADSHOTS (LE ROUX)
overuse, and shifting rainfall patterns the batch system is engineered to apply Icon’s ultimate goal, Le Roux says, is to
linked to climate change. According less pressure to water that’s less salty, reduce the cost of homebuilding by 50%.
to the United Nations, some 3.6 billion saving a considerable amount of energy. —Jonathan W. Rosen
people live in areas that experience It also increases the rate of fresh water
water scarcity at least one month of extraction by minimizing the build-up of
the year—and that number is likely salt on the membranes.
to exceed five billion by 2050. “Glob- Warsinger’s lab at Purdue, where
ally, we are truly tapping out our water he’s now a professor of mechani-
resources,” Warsinger says. cal engineering, has since worked to
Yet desalination today has major refine the batch design. His team has
limitations. Traditional reverse osmosis, developed a trailer-sized prototype it
GUTTER CREDIT HERE
in which pressurized water is forced hopes to use for pilot plants in Peru and
through a salt-removing membrane, Kenya. —Jonathan W. Rosen

Humanitarians 37
massive unmet need. The World Health

Organization estimates that there are 30
million people with amputated limbs in
poor countries, and only 5% of them have
access to prosthetics. Fitting children
with high-quality devices is particularly
expensive because they’re constantly
growing. But without prosthetics, stigma
and mobility problems keep large num-
bers of them from attending school, set-
ting many up for lifelong unemployment.
“We’re not just talking about limb dif-
ferences,” Dhaouafi says. “We’re talking
about poverty, access to education, access
to health care.”
Today, Dhaouafi has a product he
believes will help make advanced artificial
limbs more accessible. His Tunisia-based
startup, Cure Bionics, is in the process of
finalizing an adjustable multi-grip bionic
arm that will sell for about $2,000—a
fraction of the cost of similar devices.
His team plans to keep costs down by
3D-printing key components and engi-
neering much of the circuitry in-house.
But this doesn’t mean they’re skimp-
ing on quality: like bionic arms developed
elsewhere, Cure’s prototype is equipped
with sensors that allow users to operate
the hand by flexing or relaxing the mus-
cles in their residual limb. The company
is also developing algorithms to help the
arm recognize the body’s electrical signals
more accurately, which will minimize reli-
ance on an orthopedist for adjustments.
At a later stage, Cure plans to introduce
a virtual-reality headset that will gamify
the physical therapy process for children.
“Instead of a doctor asking you to imag-
MOHAMED DHAOUAFI
ine picking up an apple, you’ll be using
your hand to jump between buildings like
Spider-Man,” Dhaouafi says.
Cure Bionics Dhaouafi and his colleagues are clos-
Four years ago, during a university chal- ing in on their initial product launch:
Age 28 | Country of birth: Tunisia
lenge, Mohamed Dhaouafi found out that they’ve already tested their arm with
one of his teammates’ cousins had been five Tunisian youths and will soon initi-
His company’s artificial limbs
born without upper limbs and couldn’t ate trials at three government hospitals.
are not only high-functioning but
cheap enough for people in low- afford prosthetics. An engineering stu- Ultimately Dhaouafi hopes to offer a
CREDIT HERE
GUTTERFOUNDATION
income countries. dent at the time, he’d been searching range of high-quality, affordable pros-
for a project that would have a social thetics for young people across Africa,
impact—and as he started to research the Middle East, and beyond.
OBAMA
limb loss around the world, he found a —Jonathan W. Rosen

Li has also explored other ways to teach

artificial intelligence how to spot patterns
in linguistic data. In 2014 he and his col-
leagues correlated Twitter posts with US
meteorological data to see how weather
Entrepreneurs
affected users’ mood. First he labeled 600
tweets by hand as happy, angry, sad, and
so on. He used this labeled data to train
a neural network to assess the mood of
a tweet and cross-referenced that mood
against geolocation data for about 2% of
Their technological innovations all the tweets published in 2010 and 2011.
His results were not surprising.
bust up the status quo and lead to Moods worsened when it rained; peo-
ple expressed anger when it was hot.
new ways of doing business. But for Li it was a lesson in how hidden
information could be extracted from large
amounts of text.
After finishing his studies in 2017, he
moved back to Beijing and founded an
JIWEI LI
NLP startup called Shannon.ai, which
now has dozens of employees and $20
million in funding from venture capi-
Shannon.ai & Zhejiang University talists. Li’s company is building on the
often overplayed the importance of prox- pattern-matching work demonstrated
imity, leading to obvious mistakes. Li’s in the Twitter weather study to develop
machine-learning algorithms find the machine-learning algorithms that extract
In the last few months, Google and
grammatical structure of a sentence to economic forecasts from texts including
Facebook have both released new
chatbots. Jiwei Li’s techniques are get a much more reliable sense of the business reports and social-media posts.
at the heart of both. meaning. They have become a corner- Li has also applied deep reinforcement
stone of many NLP systems. learning to the challenge of generating
Jiwei Li applies deep reinforcement Li grew up in China and studied biol- natural language. For him it is the obvi-
learning—a relatively new technique in ogy at Peking University before moving ous next step. Once you have learned
which neural networks learn by trial and to the US, where he began a PhD in bio- to read, you can learn to write, he says.
error—to natural-language processing physics at Cornell. But he soon switched Even the best chatbots still make obvi-
(NLP), the field of computer science in fields, turning to NLP first at Carnegie ously stupid mistakes, spewing out non
which programs are made to manipulate Mellon and then at Stanford, where he sequiturs or displaying a lack of basic
human languages. became the first student ever to obtain common knowledge about the world.
By using deep reinforcement learning a computer science PhD in less than The longer a conversation, the harder it
to identify syntactic structures within three years. is for an AI to keep track of what’s been
large pieces of text, Li made machines said. Li’s techniques give AI a good grasp
better at extracting semantic information of linguistic structure. In a conversation,
from them. Syntax refers to the grammat- Li’s machine-learning keeping track of subjects and objects
ical relationship between words, while
semantics refers to their meaning.
algorithms find the is easier if the syntax of utterances is
explicit. For example, given the question
In written language, words with a grammatical structure “Shall we get started?” a bot might answer
close semantic relationship are not always of a sentence to get “Of course!”—but that response could
close together on the page. A verb and
its object can be separated by a string
a much more reliable follow any question. Li’s technique can
instead give responses more like “Yes.
of adjectives or a subordinate clause, sense of its meaning. We’ve got a lot of work to do here,” refer-
for example. Previous attempts at get- encing the content of the original query.
ting machines to parse natural language —Will Douglas Heaven
JA20_35-entrepreneurs.indd 38 6/4/20 4:32 PM

Entrepreneurs 39

ATIMA LUI residential power than a central power

Nudemeter plant, particularly when coupled with
home solar panels and batteries.
A thermionic converter consists of
a pair of metal plates, separated by a
She’s using technology to correct the
vacuum. Heat—from, say, the flame
cosmetics industry’s bias toward light skin.
of a furnace—agitates and excites the
electrons on one plate to the point that
Atima Lui grew up in long-running assump- colors that work with they leap across the gap to the cool one,
Kansas as the descen- tion that “nude” means their skin. generating an electric current. In one
dant of American white or light. Lui has managed application, Modern Electron has rolled
slaves and the daughter Lui is now deploy- to build a business the metal plates into a tube that resem-
of a Sudanese refugee, ing an AI-based app around Nudemeter, but bles a light-saber handle and fits over a
and she remembers called Nudemeter to her goals go beyond gas burner.
trying on makeup try to fix that problem. the technology itself. Homeowners could rely on rooftop
with a friend for the Through photos and Growing up, she solar panels much of the time, turning
first time as a child. a short quiz, it deter- says, she was shaped to Modern Electron’s system during the
Her friend had lighter mines a user’s skin and hurt by society’s night, on cloudy days, or in the winter
skin. “As soon as she color, accounts for how assumptions about months. If adopted widely, the product
put it on my face,” the skin is illuminated, “who gets to be an could reduce our reliance on electric-
Lui says, “there was predicts changes in entrepreneur, or who ity from centralized coal or natural-gas
nothing we could do skin tone through the gets to be a technolo- plants, which waste vast amounts of
to make it look good.” year, and helps con- gist.” That’s something energy between burning fuels and deliv-
She’d discovered the sumers of any com- else she’s trying to fix. ering power over hundreds of miles of
cosmetic industry’s plexion choose makeup —Abby Ohlheiser transmission lines. That, in turn, could
reduce greenhouse-gas emissions from
the power sector, Pan says.
The company’s technology also works
with other fuels. So if residential heating
systems eventually shift toward low- or
TONY PAN
the Seattle startup has made a new type zero-emissions sources like hydrogen,
of thermionic converter, a heat engine a change some companies and regions
first developed in the 1950s, that’s more are exploring, the thermionic converter
Modern Electron efficient than the old model at turning could make a bigger dent in pollution.
heat into electricity. Pan believes his device could have an
Age 34 | Country of birth: Taiwan
Cofounder and CEO Tony Pan believes even bigger impact in developing coun-
his company can use the technology to tries. Enabling communities to set up
His company revamps an old
convert home boilers and furnaces, which their own mini power plants would allow
device to allow you to generate
electricity in your own home. generally use natural gas or oil to heat them to skip the massive investments of
water and homes, into mini residential money and time required to build central-
Modern Electron has applied a modern power plants that produce electricity on ized generation and distribution systems.
HERECOURTESY PHOTO (PAN)
twist to an old technology. By using com- site. He says this would be a far cheaper That could bring electrification faster to
puter simulations and novel materials, and more efficient way of generating rural areas. —James Temple
CREDIT (LUI);
If adopted widely, Pan’s product could reduce our reliance

GUTTER SOONG
on electricity from centralized coal or natural-gas plants.

ASHLEY

TECH NEWS
Get our free
newsletters delivered
to your inbox.
YOU NEED. The Download
WHEN YOU
Navigate the world of tech news.
The Algorithm
Artificial intelligence, demystified.
WANT IT. Coronavirus

Tech Report
WHERE YOU
How Covid-19 is changing
our world.
fw: Economy
Your guide to growth and
WANT IT.
prosperity in the age of
technology.
Weekend Reads
Technology in perspective.
Stay in the know:

technologyreview.com/inbox
MJ20 Newsletters D3.indd 1 4/9/20 2:52 PM

42 Slug here
ENTER CAN CANADA SAVE US
THE FROM SILICON VALLEY’S WORST IMPULSES?
NARWHAL
JA20_Canada.indd 42 6/5/20 10:41 AM
Ill
43
I
t is a frigid February day when I visit
Communitech, a bustling tech hub that
occupies a renovated 19th-century tan-
nery in the city of Kitchener, Ontario.
Inside the brick-and-beam space,
Harleen Kaur opens her phone and
pulls up her latest creation—an attempt to tackle the
problem of misinformation. It’s an app called Ground
News, a combination news aggregator and social-media
platform that combats falsehoods with the help of AI
and on-the-ground verification carried out by its users.
Tapping on the headline “Buttigieg slams Trump:
My marriage never involved sending ‘hush money to
a porn star,’” I learn that over the previous two days
the story had been covered by 14 outlets, with a collec-
tive “lean left” bias, according to the app. Had I been
at the town hall meeting where Pete Buttigieg, then a
presidential candidate, made this remark, I could have
created my own story about it using the app’s Citizen
Journalism feature (the part intended as a check on
fake news), which other users would then have been
able to contest or confirm.
Kaur, an aerospace engineer turned serial tech entre-
preneur, was living in the US when the idea for Ground
News hit her like Newton’s apple. Though motivated
by a growing problem in America—this was 2016—she
decided to return home to Canada to build the new
company. Venture capital and other investment flows
fast and free south of the border, she says, but it was
more important to embark on the next phase of her
career where “Canadian values” reign.
“Canada is more measured, more considerate. Our
value system is not just making money and being suc-
cessful,” says Kaur, who moved as a girl from India
to Brampton, an immigrant enclave in the suburbs
of Toronto. “Canadians are nice to each other. I think
being nice matters. Nice has value.”
Kaur is not the only one drawn away from the US
by Canadian benevolence. In recent years the country
has become a magnet for technology talent, reeling
Canadians back home and diverting the stream of over-
seas applicants away from Silicon Valley to Montreal,
Vancouver, and the Toronto-Kitchener-Waterloo cor-
ridor. These are areas long known for incubating and
exporting innovation—from Research in Motion, the
company founded above a Waterloo bagel shop that
ushered in the smartphone age and later took the name
By BRIAN BARTH
Illustrations by David Biskup
of its flagship product, the BlackBerry, to the neural
networks of Geoffrey Hinton, the University of Toronto
professor whose AI company was acquired by Google
in 2013. Canadian media have christened the region
“Silicon Valley North.”
JA20_Canada.indd 43 6/5/20 10:41 AM

Some are drawn by the they dropped below 70% at 12 Yung Wu is the CEO of the Canada produces narwhals.
image of Canada as a liberal US tech firms—while the wait MaRS Discovery District, a But, Wu says, the comparison
utopia, where diversity, inclu- time went from five months block-size campus in down- goes beyond their financial val-
sion, and humility triumph to nearly 10. The number of town Toronto where firms can uation. “In the Valley you find
over greed and bigotry. While H-1B applicants, after rising for rent space, mingle in a massive this chase for imaginary ani-
this branding may have been years, declined after Trump was central atrium, and tap into ser- mals. The narwhal is actually
carried to excess by Prime elected, from 236,000 in 2016 to vices designed to help startups a real thing,” he says. “It’s not
Minister Justin Trudeau, the 199,000 in 2017. The equivalent and “scale-ups” grow. He has propped up by private valua-
allure is real. Canadian visa program, mean- seen the revenue of its 1,500 tions that are intended to raise
As President Donald Trump while, approves 95% of appli- companies almost triple in the the round on the last private
tightens the US border—in cations in two weeks or less. past two years, but he insists valuation, with no resemblance
April he placed a 60-day ban As a result, Toronto added that Canadian tech is on a qual- necessarily to a real company
on most green cards, ostensi- more new tech jobs between itatively different path from its that serves real customers with
bly to protect American jobs 2013 and 2018 than any other US counterpart. “I don’t think real revenue. A narwhal is rare,
as covid-19 tanked the econ- North American market sur- the bro culture would have but it’s not an imaginary thing.”
omy—Trudeau opens Canada’s veyed. It is now ranked behind really developed in the same The question is whether
arms ever wider. In 2018, he only San Francisco and Seattle way over here, for instance,” he Canada’s quiet narwhals can
pledged to admit an additional for tech talent by the real estate says. Canadian values may play make a big enough splash to
40,000 immigrants over the giant CBRE. Invest in Canada, a role in that, but demographic change the trajectory of the
next three years, raising the a federal agency charged with differences are also part of the global tech industry.
quota to 350,000 by 2021, and attracting global firms to set equation—Toronto is consid-
covid hasn’t changed that pol- up shop, advertises Toronto as ered one of the most diverse
icy: “Immigration will abso- having the “highest concentra- cities on earth, and more than “A COLONIZING EXPERIMENT”
lutely be key to our success tion of AI startups in the world.” 50% of its residents were born Canadians are notoriously
and our economic recovery,” The government has gone so in another country. Likewise, at polite and generally go out of
Canada’s immigration minister, far as to pay for billboards in MaRS, which bills itself as the their way to not criticize their
Marco Mendicino, said in May. Silicon Valley that read “H-1B “largest urban innovation hub” southern neighbors. Despite
While US immigration pol- Problems? Pivot to Canada,” in North America, more than this, there is a burning nation-
icy has been tough on many with a link to the nation’s immi- half of all company founders alism deep down that can take
industries, the issue is especially gration website. were born abroad. the form of disgust toward
acute in the tech sector, which Kaur thinks this “niceness” Canada has its own techno- many aspects of American cul-
relies on highly skilled foreign also boosts her bottom line. mythology. Instead of tech bros, ture—such as an excess of indi-
workers on both sides of the “Having ‘Brand Canada’ asso- it has a workforce portrayed as vidualism and self-important
border. In the US, approvals ciated with us is a benefit,” she diverse, reserved, and polite. verbiage expressed at a high
for H-1B visas, the kind typ- says. “We have a halo effect Where Silicon Valley prizes volume. And sometimes that
ically given to skilled tech around us of being trusted and trend-setting consumer prod- disgust erupts like projectile
workers, dropped from 94% neutral.” ucts, Toronto’s startups tend to vomit. This was the case when
of applications in 2015 to 76% be more focused on services Sidewalk Labs came to town.
in 2019—one study found that and products for business and In March 2017, Waterfront
government clients that are Toronto, a government agency
less likely to capture the public charged with redeveloping a
imagination. Where California 2,000-acre (800-hectare) strip
produces “unicorns,” private of former industrial land along
58,0000
58,00
tech companies valued at a bil- the shores of Lake Ontario,
lion dollars or more, Canadian asked for proposals to build a
techies speak of building “nar- smart-city district on a 12-acre
whals”—named for the small, parcel known as Quayside.
reclusive whales, with long Trudeau presided over a lav-
tusks sprouting from their ish ceremony that October to
heads, that roam Arctic waters. announce the winner: Sidewalk
NET TECH JOBS ADDED IN TORONTO The US produces far more Labs, a New York–based urban
FROM 2013 TO 2018 unicorns per capita than innovation company owned
JA20_Canada.indd 44 6/8/20 1:04 PM

Enter the narwhal 45
“THE OPPOSITE OF FACEBOOK”: HOW CANADA THINKS ABOUT TECHNOLOGY
HARLEEN KAUR YUNG WU JOHN RUFFOLO KURTIS MCBRIDE IAIN KLUGMAN

GROUND NEWS MaRs ArcTern Miovision Communitech
Ventures
“Our value “I don’t “Monopoly “That’s the “There is a
system is not think the bro power over thing that different
just making culture would data … This Canada could social contract
money and being have really is something bring to the in this
successful. developed in that we need world—replacing country. We’re
Canadians are the same way to defend the data collaborators.
nice to each over here.” ourselves monopolies That’s our
other. I think against.” with a data secret weapon.”
being nice has collective.”
value.”
by Alphabet, Google’s parent. then be exported globally, private information (location, Google executive who had also
Sidewalk had been chosen over extending Alphabet’s market purchasing habits, and so on) been called to testify. “History
the Canadian companies that dominance from cyberspace into Google’s data-hungry maw. offers sobering lessons about
had applied, but the premier of to public space. It would be a Data sovereignty—the idea societies that practice mass
Ontario, the mayor of Toronto, crowning achievement not just that a nation’s data should be surveillance.”
and Alphabet’s then chairman, for Sidewalk, but for Alphabet held on servers within its bor- A deluge of negative head-
Eric Schmidt, spoke glowingly and the Canadian government. ders, governed by its laws and lines dogged Sidewalk Labs
of the plan to build a neighbor- Some Canadian-bred tech- thus ultimately by its values— throughout its first year. City
hood “from the internet up.” nologists, however, were less has gained traction around councilors and members of
Soon there was talk of than impressed. Jim Balsillie, the world. The EU’s General parliament began speaking
expanding the development the billionaire who had been Data Protection Regulation out against the project. As the
to the Port Lands, 800 acres co-CEO of Research in Motion (GDPR), which went into scope of the company’s plans
of derelict industrial prop- until 2012, launched an ad hoc effect in 2018, is based largely for data collection became
erty adjacent to Quayside. It campaign to quash the project. on this principle, prying con- clear, the former privacy com-
was arguably the most ambi- Quayside “is not a smart trol of the information col- missioner of Ontario, who had
tious smart-city initiative in city,” he wrote in an op-ed. lected on its citizens out of been retained by Sidewalk Labs
the world, including plans to “It is a colonizing experiment the hands of the tech giants as a consultant, resigned. “I
use sensors and monitoring in surveillance capitalism.” and the American laws that wanted this to become a smart
to create a vast amount of data In Balsillie’s view, the net- govern them. Balsillie urged city of privacy—not a smart
that could be used to serve the work of sensors proposed for Canadian lawmakers to follow city of surveillance,” she said.
needs of households and work- Quayside—which Sidewalk suit. “Facebook and Google are In October 2019, Waterfront
places, aid with transportation, Labs said were necessary to run companies built exclusively on Toronto unveiled a revised
and even charge citizens by robotic trash collection, high- the principle of mass surveil- agreement with Sidewalk
COURTESY PHOTOS
the item for their trash. The efficiency utility systems, and lance,” he told the Canadian Labs that put the company
idea was to develop the digi- other digital improvements— parliament at a 2018 hearing on on a much shorter leash. The
tal architecture for an urban amounted to an Orwellian the Cambridge Analytica scan- scope was limited to the orig-
operating system that could power play, feeding Canadians’ dal, where he sparred with a inal 12 acres, not the coveted
800, and data collection would has hosted a startup incubator the area to discard the “Silicon things” (“Maybe it’s time to
remain under the control of in the building since 2013. Valley North” label and assert slow down and fix things,” he
the government, not the com- In the 1980s, the twin cities a uniquely Canadian identity. says) feels timely as the world
pany. Then, in May of this year, of Kitchener and Waterloo, not “The Valley was founded by ponders alternative futures.
Sidewalk Labs announced it far from Detroit, were consid- deviants—we call them liber- But even though Canada
was pulling out. The company ered part of Canada’s Rust Belt, tarians now—who didn’t have wants to capitalize on its
cited the pandemic economy a region littered with shuttered respect for regulations,” he says. non-Valley identity, what
as the reason for the deci- factories lost to the vagaries of “That’s why they dream up ideas exactly is the alternative on
sion, while detractors framed globalization. One bright spot like Uber and Airbnb. There is a offer? “Being the opposite of
it as a convenient excuse for was the University of Waterloo, different social contract in this Facebook, basically,” Klugman
Sidewalk to slither away with- where the computer engineer- country. We’re collaborators. says. “It’s not just can you build
out losing face. “This is a major ing program was growing in That’s our secret weapon.” it, but should you build it?
victory for the responsible cit- renown. That’s where a Greek- Klugman, a compact man And being responsible for the
izens who fought to protect Turkish student named Mike with decidedly non-Canadian implications.”
Canada’s democracy, civil and Lazaridis was studying before intensity (ironically, this cham- That’s easy to say—but
digital rights,” Balsillie told the he dropped out in 1984 to start pion of north-of-the-border of course Google, Facebook,
Associated Press. Research in Motion.
Balsillie declined to be Improbably, a handful of
interviewed for this article, other successful technology
but John Ruffolo, a prominent businesses emerged in the area,
venture capitalist in Toronto including OpenText—which
who is a longtime ally of his, makes information manage-
told me it had been a struggle ment software for large com-
“to smack into people’s heads panies—and, more recently,
in the government” the dan- the messaging app Kik. The
gers of Facebook’s and Google’s founders of these firms created
“monopoly power over data.” Communitech out of necessity:
But he believes Canada’s far from other concentrations of
tech industry activists now have capital and innovation (Toronto
lawmakers’ attention: “If you was not the financial center Waterloo, “Canada’s
Silicon Valley.”
think you’re going to get con- that it is today), they relied on
trol over public infrastructure each other for support. It was a
where I, as a private citizen, radically different genesis from values is originally from and the rest all started with
might be subject to facial recog- that of Silicon Valley, with its Colorado), delivers this sermon noble, collaborative visions of
nition walking on a public side- cutthroat culture and its roots as he leads me past a banner their own. What starry-eyed
walk or some other intrusion in the military-industrial com- advertising a $1 million prize Silicon Valley entrepreneur
of my privacy—this is some- plex rather than working-class for the company with the most didn’t speak the same way
thing that we need to defend immigrant communities. promising AI solution to fake in 1995, or 2000, or 2005?
ourselves against, because we Essentially a membership news. We go into a conference Perhaps Canada’s lack of suc-
will slip into tyranny, just like organization, Communitech room, where he lays out his cess has made it easier to stay
where I think China is today.” has seen its original list of vision for how Canada’s brand aloof: while America’s techno-
23 companies grow to more of tech will quietly take over optimists have had ample
than 1,400 and has spawned the world. opportunity to sell their souls
“FOUNDED BY DEVIANTS” a national network of 29 hubs After years of sky-high pub- to the devil as their share prices
Communitech, the Kitchener operating on the same model. lic approval for the industry, the soared to dizzying heights,
tech hub, is ground zero of the With hundreds of new technol- so-called “techlash” provides few of their Canadian coun-
country’s tech scene. Founded ogy firms appearing each year, an opening, Klugman believes, terparts have had so far to fall.
WEI FANG /GETTY IMAGES
in 1997, it has grown into some- Waterloo boasts the highest for “collaborative capitalism” to And angelic reputation aside,
thing of a national legend, a startup density on the planet take root. Even if it is oppor- Canadians are not immune to
place where tiny startups rub after Silicon Valley. tunistic boosterism, the notion ethical lapses: in fact, some of
elbows with Google execu- Iain Klugman, the CEO of that Canada offers an alterna- the most-hyped narwhals have
tives—the Silicon Valley giant Communitech, says it’s time for tive to “move fast and break had their share of controversy.

Enter the narwhal 47
37,000 30,000
“I think there’s a real oppor-
tunity for Canada to lead this
next evolution of technology,”
she says. “There’s so many
unanswered questions to define
a decidedly Canadian way, but
whatever that is, it needs to
DECLINE IN ANNUAL APPLICANTS IN 18 come from a sense of generos-
ity, of understanding the chal-
H-1B APPLICATIONS IN MONTHS TO MOBSQUAD,
lenges that people are having
THE US IN THE FIRST WHICH PROMISES
today, and their need to lead
YEAR OF THE TRUMP CANADIAN VISAS TO meaningful lives.”
ADMINISTRATION THOSE OFFERED A JOB One person working on what
a “decidedly Canadian way”
might look like, at least for smart
cities, is Kurtis McBride, CEO of
In the 2000s, Research in frame, many other American Meanwhile, the Canadian Miovision, which specializes in
Motion was embroiled in a hubs were losing them, includ- government has made jabs at high-tech traffic management.
stock options scandal. Kik’s ing New York (9,000 net jobs the Valley’s business model, McBride had been part of
messaging app boasted 300 lost), Raleigh-Durham (10,000), including a new law that a group advising Waterfront
million users when the com- and Boston (34,000). requires internet platforms to Toronto on the Quayside proj-
pany abruptly shut it down Sometimes this flow is very track and publish the identity ect; he’s also the board chair of
last year amid a financial clear and deliberate. For exam- of anyone buying politically the Open City Network, which,
scandal and complaints that ple, H-1B visa holders who have oriented advertising. (Google, pre-pandemic, was developing
it enabled pedophiles to stalk been furloughed or laid off—as claiming the regulation was too standards and digital architec-
its predominantly teenage many American tech workers onerous, opted not to run elec- ture for smart cities, including
users. There is little to sug- have been in recent months— tion ads as a result.) guidelines for data governance.
gest that a small Canadian tech have 60 days to leave the US. In 2019, the Trudeau (It has now pivoted to build-
company wouldn’t turn into a “Those people have to government unveiled a 10 ing a platform on which the
monopolistic beast if it grew go somewhere,” says Irfhan Commandments–style “digi- public and private sectors can
big enough. Rawji, the CEO of MobSquad, tal charter,” which proclaims share data.)
a Calgary company he founded Canadians’ right to control their Keeping everything open-
in 2018 to siphon such workers data. The government has not, source, he told me, made
“THEY HAVE TO GO SOMEWHERE” to Canada. however, enacted legislation it monopoly-resistant, the
Still, that optimistic and antag- MobSquad hires H-1B hold- to give teeth to those procla- opposite of the Sidewalk
onistic position is proving a ers who were unable to renew mations, à la the EU’s GDPR. Labs approach of giving con-
magnet for talent. Canadians their visas, sets them up with But as a country of 36 million trol of digital architecture to
have complained of a brain a Canadian equivalent, gives with a GDP far smaller than the world’s biggest internet
drain for decades—the best each one a desk in its Calgary California’s, Canada could company.
software engineers, actors, and coworking space, and con- hardly be expected to lead on “I think there’s an oppor-
comedians always seem to go tracts them back to their orig- regulatory reform. Perhaps if tunity to essentially write the
south—but every tech execu- inal US employers as virtual it has a role to play in reining rules of an economy that would
tive I spoke to proudly reported workers. The MobSquad web- in the industry’s excesses, it govern how data generated from
that the flow, at least in their site, as a solution to America’s may be through exerting soft public places and physical
industry, has reversed. Between “software engineering talent power—leading by example. spaces is used,” McBride said.
2013 and 2018, Toronto alone crisis,” advertises Canadian Ana Serrano, a member of John Ruffolo, the venture
saw a net gain of nearly 58,000 visas within four weeks and the anti–Sidewalk Labs coalition capitalist and staunch Sidewalk
tech workers, more than any citizenship within four years. Block Sidewalk, thinks Canada’s critic, put it to me more
other North American city sur- Rawji says he had more than alternative vision—not what simply: “Canada could become
veyed. While San Francisco and 30,000 applicants in the first it is against, but viable busi- the Switzerland of privacy.”
Seattle continued to add tech- 18 months, and now “we are ness models that show what it Brian Barth is a freelance
nology jobs during that time busier than we’ve ever been.” is for—is still in its fetal stage. journalist based in Toronto.

Companies on the internet

The Great Depression offers lessons for how to give ordinary can collect data about people’s
people a say in the economic recovery from covid-19. behavior in ways old phone com-
By Nathan Schneider / Illustration by Sophy Hollington
panies and mail carriers never
could: a telecom can’t listen to
your phone conversations and
A CRISIS IS
send you relevant robocalls.
Ride-sharing apps got their start
in part by bypassing regulations
their taxicab competitors had
NO EXCUSE NOT TO
to follow. Gig-economy plat-
forms routinely claim the right
to ignore hard-won labor protec-
tions on the grounds that they
REGULATE TECH
offer part-time freelance work,
even though in many cases this
work involves the kind of control
over workers that is tantamount
to standard employment.
There has long been a pre-
sumption in some quarters that
the old rules don’t apply to new
tech. Earlier this year, before the
virus set in, Michael O’Rielly, a
The “techlash,” allegedly, is over. commissioner at the US Federal
An April op-ed in the San Jose Mercury News, Silicon Valley’s Communications Commission,
local paper, put it most directly: “Covid-19 response will end all the spoke at the university where I
Big Tech bashing.” An article published by the Brookings Institution teach. He expressed his hope
later that month echoed the new received wisdom: “Prior concerns that with the days of “circuit-
Nathan Schneider
about the industry’s market power, privacy practices, and content is a professor of switched copper networks”
moderation policies—all of which posed a major challenge just media studies at behind us, the FCC’s role would
the University
months ago—no longer enjoy the same political salience.” of Colorado,
“diminish exponentially,” like
The argument is that covid-19 has taught us to stop worrying and Boulder, where he “a puff of smoke on a windy
directs the Media day.” But we find ourselves in a
love Silicon Valley—to simply embrace the connections it brings to Enterprise Design
our quarantine and the surveillance it can apply to contact tracing. Lab. moment when the companies
But as people find themselves relying on the tech economy in fuller, the FCC regulates mediate more
of our lives than ever before.
more intimate ways, they are finding new reasons to be concerned.
Indeed, many of the US’s
An Amazon vice president stepped down in May in support of
major antitrust laws were cre-
workers who were fired for organizing for better workplace safety
ated for crises not so unlike the
measures against the coronavirus. Low-wage workers from other one we face today—times of
companies, including Instacart, Target, and Walmart, have gone super-powerful magnates and
on strike for similar reasons. Airbnb hosts are disgruntled that the widespread economic upheaval.
platform they work for and lobby for is giving customers who cancel These laws, crafted for the
bookings full refunds, leaving hosts with no income and all the costs. railroads and Standard Oil,
In moments of crisis, when new technology seems to offer quick empower regulators to, among
and easy answers, it might appear difficult to devise an imaginative other things, break up any com-
COURTESY OF THE AUTHOR
response to the large tech firms’ growing power. But even though the pany abusing its market dom-
litany of things that tech platforms get away with is quite remark- inance. Regulators have not
able, tools for fixing some of tech’s deepest problems are closer at recently exercised these pow-
hand than one might think. ers against Big Tech because
for decades they have narrowly
JA20_Schneider_new_deal.indd 48 6/5/20 12:57 PM

Essay 49
electric companies could be

used to bring customer-owned
broadband to underserved com-
munities. Some old rural electric
co-ops are offering fiber-to-the-
home already.
Furthermore, gig workers
and customers who rely on
them currently have to use
investor-owned platforms. But
one proposed bill in California,
the Cooperative Economy Act,
would enable platform workers
to organize co-ops that could
collectively negotiate with plat-
forms—and perhaps even build
platforms of their own. This
would enable these workers,
many of whom are now essen-
tial as drivers and delivery peo-
ple, to obtain better wages and
working conditions.
Quarantine and remote work
fixated on consumer prices as from grocery delivery to elder farm country, where investor- also leave many people more
the measure of whether a market care, they deserve every protec- owned utilities hadn’t bothered dependent than ever on com-
is being monopolized—a mea- tion society can reasonably offer. to string lines. Low-interest munication platforms, which
sure that doesn’t work for ser- Regulations alone, however, loans through the Department of typically collect personal data
vices, like Facebook and Google, are not enough. Policy should Agriculture enabled communi- for uncertain purposes. This
that are free. This would change enable more than it prevents. ties to organize cooperatives— shouldn’t be a necessary trade-
if regulators allowed themselves In the 1920s and 1930s, US leg- nearly 900 of which still operate off. Using free, open-source tools
to see how far-reaching the old islators put this principle into today. The loan program now like NextCloud for file-sharing
antitrust mandate against mar- practice. Following the 1929 earns more than it costs. Like the and Jitsi for videoconferencing,
ket manipulation really is. With stock market crash, it was clear housing policies of the time that groups can manage their own
many smaller businesses now on that banks were not account- gave us the 30-year mortgage, it privacy-protecting systems and
the brink of collapse, the dan- able to their clients, and there was a public policy that enabled decide for themselves how their
ger of consolidation has never were huge swaths of the coun- widespread private ownership. data is used. Public investment
been greater. A moratorium on try that banks didn’t serve. In These were some of the most in projects like this could ensure
mergers is probably a necessary addition to new regulations powerful economic develop- that, as with credit unions, peo-
stopgap. that constrained the banks, the ment programs in US history. ple have the means to organize
There’s a similar story of 1934 Federal Credit Union Act They introduced dynamism and alternatives when the big plat-
amnesia in labor law. The turned a few local experiments decentralization to markets in forms aren’t meeting their needs
gig-economy platforms have in community finance into a danger of being held in thrall or respecting their values.
all but admitted that their government-insured system. to monopoly and exploitation. If The internet may have
business depends on system- Member-owned, member- we want a more inclusive tech near-magical powers that can
atically violating labor protec- governed credit unions prolifer- economy, the New Deal legacy help us get through the coro-
tions. California recently woke ated. They held banks to higher would be a good place to start. navirus crisis, but making tech-
up to that fact, passing a law standards and brought financial Internet users need the nology firms accountable can
reclassifying many gig workers services to places where there capacity to form cooperative begin with lessons learned from
as employees. Especially now, had been none. alternatives to the dominant the last depression. Good tech
when people with precarious In similar fashion, two years platforms and infrastructure. policy requires recognizing that
incomes are risking their health later, the Rural Electrification For instance, much the same tech is just another way of wield-
by providing essential services Act helped bring electricity to model as that of the cooperative ing power.
JA20_Schneider_new_deal.indd 49 6/4/20 1:56 PM

JA20_MacBride.VC.indd 50 6/8/20 4:36 PM
G A
M E
6/8/20 4:36 PM
51
Elizabeth MacBride
Nicolás Ortega
Illustrations by
By
Venture capital has become extremely successful at creating a certain kind of innovation—
but the pandemic has exposed its broader failures.
LOSING THE WINNERS’
JA20_MacBride.VC.indd 51
I FELT BAD
asking Zack Gray to repeat his story. He
was used to it, he said. It’s the founding
tale of his startup, Ophelia; he’d already
told part of it in his commencement speech
at Wharton, and to potential investors.
Venture
“There was a girl in my life,” he started. “I call her my girlfriend. We capital has
met when I was 14.” They dated, on and off, and stayed friends.
She was one of a generation who slid into opioid addiction through
been the
painkillers. A user for five years, she had the means to seek treatment engine of US
innovation for
after her addiction grew, but she didn’t want rehab or therapy.
Then, last spring, the call came: she

had overdosed. By the time Gray got to
capital firms in the US had risen from 946
in 2007 to 1,328 in 2019, and the amount years.
the hospital, she was gone.
“I just started thinking, ‘What could I
of money they were managing had swol-
len from $170.6 billion in 2005 to $444 But investors
have done to prevent this?’” he said.
To answer that question, he researched.
billion in 2019.
Not all the numbers were so positive, are finding
Since he was finishing up his MBA, the
approach that seemed obvious was to
however. This largely white, largely male
corner of finance has backed software fewer ideas
build some kind of business or service.
What if his friend had been able to get
companies that grow fast and generate
large amounts of money for a shrinking that fit their
medicine to treat the chemical condition
of addiction, without the embarrassment
number of Americans—companies like
Google, Facebook, Uber, and Airbnb. But preferred
and hassle of group therapy? Would insur-
ance companies buy in to his concept?
Could he build a big company to help a lot
they don’t create many jobs for ordinary
people, especially compared with the
companies or industries they disrupt. And
pattern.
of people like her? He’d need investors to things have been slowing down. Recently,
believe in the idea. venture capitalists have found fewer and
As I listened to Gray explain what fewer ideas that fit their preferred pattern.
he was doing, headlines were washing By the end of 2019, the industry had $121
over America. “A Torrent of Job Losses billion in “dry powder,” money in search
Threatens to Overwhelm the US Economy.” of an entrepreneur or idea to invest in. I
“Doctors Say Shortage of Protective Gear wanted to know what was going on.
Is Dire.” “Coronavirus Hitting Homeless As covid-19 took hold of the world, my
Population.” plans to meet Gray and his peers changed.
It didn’t seem like a great time to be And suddenly, the questions became more
raising money. urgent. Was venture capital producing the
I had originally planned to meet Gray kinds of inventions society needs? Sure,
in person. I was scheduled to fly out to when we have to (or want to) stay home,
California in March to attend the startup Zoom helps us work remotely, DoorDash
accelerator Y Combinator’s famous Demo keeps us fed, and Netflix gives us some-
Day. The event would host 1,000 investors thing to watch. But where was the cure, or
and introduce them to nearly 200 vetted the better protective gear, and why hadn’t
and prepped startups from around the venture capital—the financial engine of
globe. Ophelia was one of those startups. innovation—funded those ideas?
I was going to Demo Day because ven- In the 1950s and ’60s, technology took
ture capital had been America’s financial us to space. In the 1980s and ’90s, tech-
engine of innovation for years, and I wanted nology helped spread democracy. Now
to see if that was still true. Many stats our national mission was ... to be able to
suggested it was: the number of venture never leave the house?

Losing the winners’ game 53
more on R&D than the rest of the world

combined. While that fire hose of cash
flowed, the first venture capitalists found
many winners to bankroll.
The link to government is still very
much there in today’s technology com-
panies. Google’s early work came out of
the Clinton-era Digital Libraries project
at Stanford, and the CIA was Palantir’s
first customer in 2003—and its only one
until 2008.
O’Mara says there isn’t anything wrong
with tech companies’ being built through
US research dollars. In fact, she argues, the
most important decision of that era was for
the government to pour money in without
exerting too much control. But, she adds,
a mythology has grown up that focuses on
lone heroes and rule breakers rather than
the underlying reasons for a company’s
or technology’s success. “Hooray for the
internet that it’s still cranking,” she says.
“But you did not do this by yourself.”
In 2011, one of the bigger cowboys
of venture capital, Marc Andreessen—
the Netscape cofounder who now runs
Andreessen Horowitz, one of Silicon
Valley’s most influential investment
firms—wrote a famous essay titled “Why
What capital wants take most of the returns. It’s a hard, fast Software Is Eating the World,” in which he
When I want to understand finance, I process: to get even one or two big winners described the destruction of middle-class
call my friend Charley Ellis. He was on within the standard 10-year time frame, jobs in America and predicted the venture
the boards of the investment manage- a venture capital fund invests in dozens profits of the following decade.
ment company Vanguard and the Yale of startups. Most companies that don’t He was right: software companies are
Endowment, and he wrote a bible for grow fast enough get no more invest- attractive to investors because they can
investors called Winning the Loser’s Game. ment and die. generate large returns, often by replacing
“The fact that it’s called venture capital Venture capitalists sell themselves as people in industries those software firms
is a terrible distraction,” he told me. “It’s the top of the heap in Silicon Valley. They come to dominate—for example, travel
really human resources.” are the talent spotters, the cowboys, the agents, whose work is now done by flight
What he means is that successful ven- risk takers; they support people willing booking websites.
ture capitalists aren’t necessarily those to buck the system and, they say, deserve Venture capitalists look for companies
who find and fund the most innovative to be richly rewarded and lightly taxed that can reach IPO size, which means
ideas, but the ones who know how to spot for doing so. they need an idea that can find a big mar-
founders capable of building a company The image, however, doesn’t strictly ket. These factors combine to produce a
that will eventually be acquired or go match the history of the Valley, because very specific set of requirements, which
public. The $1 million that might be spent it was “the system” that got everything Y Combinator has reverse-engineered to
to buy a 10% share of an early-stage busi- started. After Sputnik launched the space great success.
ness turns into much more if that company race, the federal government poured “Investors are a simple-state machine,”
ends up being worth $10 billion. money into silicon chip companies. Michael Siebel, the accelerator’s CEO, told
Venture capital firms sell their ser- Historian Margaret O’Mara documents me. “They have simple motivations, and
vices to investors like hedge funds, pen- this well in her book The Code: In the it’s very clear the kind of companies they
sion funds, and wealthy individuals, who early 1960s, the US government spent want to see.”
JA20_MacBride.VC.indd 53 6/9/20 11:45 AM

But some of the other inputs, either accelerators like Y Combinator to find, check-ins online), and—since some 2 to
consciously or subconsciously, have been filter, and train entrepreneurs who meet 3 million people in the United States are
assumptions about the kind of person who their needs. Twice a year, thousands of addicted to opioids—it had a large poten-
can help generate outsize returns. The top startups apply to be part of its three-month tial market.
founders “all seem to be white, male nerds training program, in which they hone Y Combinator advised Gray not to tell
who’ve dropped out of Harvard or Stanford their ideas and learn to speak VC. Then, me how much funding he was seeking,
and they absolutely have no social life,” John at the carefully scripted Demo Day, they because it looks bad if you don’t hit the
Doerr of Kleiner Perkins—one of the most are introduced to venture capitalists from mark. But his idea was built to appeal to
influential investors in the Valley—noted in around the world. investors. Other ideas he’d considered ear-
2008. “So when I see that pattern coming Founded in 2005 by an earlier gen- lier were more like moonshots—hotels for
in … it was very easy to decide to invest.” eration of Silicon Valley luminaries, Y homeless people, for example.
Even as investors have found oppor- Combinator has helped launch Instacart, “The challenge here is to build a busi-
tunities dwindling, as evidenced by that Dropbox, Airbnb, and Stripe, among oth- ness that does good and can raise money.
growing stash of “dry powder,” venture cap- ers. Besides whatever they get from other You need to figure out how to monetize
ital has continued to flow almost entirely investors, it gives each company $150,000 it,” Gray said. “If you can help people and
to the same kinds of male founders. Only in exchange for a 7% ownership stake. they can pay for it, that’s the key.” For all
just over 2% of VC money in the US went As of October 2019, according to Y his idealism, he had adapted to a ven-
to female founders in 2017 and 2018. Combinator, 102 of its graduates had a ture system that has evolved to act as the
Still, many people in the Valley think valuation of more than $150 million (not spear tip of profit-seeking capitalism and
this system works well. including some that didn’t want their valua- American individualism.
“If you have a terrific founder with a tions disclosed). Those companies, worth a I asked Charley Ellis why he thought all
terrific idea, they’re going to get funded,” combined $155 billion, have created 50,000 these smart investors and entrepreneurs
one investor told me. “Never has the jobs in 15 years, the accelerator says. Of hadn’t put their time and money into health
system been more efficient at getting cap- the new batch, I was drawn to Ophelia systems that could detect infectious dis-
ital to the right people.” because it was a telehealth company, and eases, or quicker ways to develop drugs
When I came out of my office after Gray seemed unusually thoughtful. and vaccines, or unemployment benefit
that particular interview, I found that my He told me he had reservations about systems that could cope with a sudden
16-year-old daughter had been listening. the venture capital model, especially right crush of applications.
“He doesn’t seem to realize he’s the Once- now. “I spent a lot of time philosophizing Ellis pointed out that people have a
ler,” she said, referring to the character in and rationalizing the moral rectitude of hard time seeing outside their universe.
Dr. Seuss’s The Lorax who thought he was what I’m doing,” he said. “People inside an industry are so focused
making a great company when really he Still, he hoped to find an investor who on creating money for their industry,” he
was destroying the environment. would help him reach 500 patients in the said. “Nobody wants to stop the game.”
first year, and many more later. Ophelia Gray is definitely in the game. He lost
Playing the game matched some criteria those investors his father, who worked on Wall Street, to
In their search for the elusive home run, typically look for: it was software-driven cancer when he was a young teenager and
venture capitalists increasingly rely on (allowing patients to do follow-up medical then went to Columbia University, where
$121 BILLION
AMOUNT OF UNINVESTED
65%
PROPORTION OF VENTURE CAPITAL
2%
AMOUNT OF US VENTURE
MONEY IN VENTURE CAPITAL FIRMS WITH NO FEMALE PARTNERS CAPITAL THAT WENT TO FEMALE
FUNDS, 2019 FOUNDERS IN 2017-18

he studied philosophy and astronomy. After Susan Choe, the founder of Katalyst
he figured out that academia moved too Ventures, is an investor in Zipline, whose
slowly for him, he enrolled at Wharton, drones deliver medical supplies in poor
“The challenge the University of Pennsylvania’s business

school. This Ivy League pedigree gave him
countries where infrastructure is lacking.
It’s valued at more than $1 billion. She also
here is to access to a world most entrepreneurs can’t

dream of reaching. Adam Grant, a famous
pointed me to All Raise, an organization
that promotes women in venture capital. It
build a UPenn management professor, became an

adviser to Ophelia and he discussed his
reported in 2019 that a record 54 women
became VC partners, though 65% of venture
business that idea with Tom McClellan, Barack Obama’s

drug czar.
capital firms still have no female partners.
“Change is being driven by the fear of
does good Listening to Gray, it was hard not to

think about the advantages wealth and con-
being left behind,” says Choe, who says
that limited partners—investors—in her
and can raise nections offer. These benefits have been

quantified by researchers who studied 1
funds include executives from outside the
US. Millennials tend to be drawn toward
money ... If million US patent holders and looked at

their parents’ income. Low-income stu-
more diverse teams, too, she says.
She is among those who make the case
you can help dents who scored in the top 5% in math

were no more likely to become inventors
that venture capital firms overlook products
and services that cater to ignored commu-
people and than below-average math students from

affluent families, they found. Meanwhile,
nities or create new markets. “Investors
are leaving money on the table, and they
they can pay if women, minorities, and children from

low-income families were to invent at
are missing innovation because the people
that are running these VCs cannot relate
for it, that’s the same rate as white men from families
with incomes in the top 20%, the rate of
to the preferences of people that are liv-
ing outside their experiences,” says Lisa
the key.” innovation in America would quadruple.

The advantages of wealth build on each
other. Information is an important one:
Green Hall, a fellow at Georgetown’s Beeck
Center for Social Impact & Innovation and
former CEO of Calvert Impact Capital. “In
Gray knew from the beginning that he the white male culture ... those cultures
wanted to get into Y Combinator, which are extremely narrow. For women and
he’d heard about as a student. And getting people of color, those cultures are much
into the accelerator, in turn, “de-risked more expansive.”
and legitimized Ophelia,” he says. With It brought to mind Jasmine Edwards,
that important stamp of approval, he was a black woman from Tampa, Florida, who
able to recruit a cofounder, Mattan Griffel, launched an education startup that aimed
a more experienced entrepreneur who to help schools with low-income students
became his chief operating officer. find better substitute teachers. With 200
substitute teachers on the platform and
Slow evolution three schools as paying customers, the
Still, while Ophelia fits the traditional startup ran out of time and cash, and it
Zack Gray profile of an investable company for the folded. What could have been different if
likes of Y Combinator and the venture she had been able to raise the funds she
capitalists who go on to fund its startups, needed to continue?
the industry has been changing, at least
a little. Recent years have brought a new What are you building?
class of “impact investors,” who eschew On April 18, Marc Andreessen emerged
the profit-obsessed venture capital model with another essay, this time occasioned
to focus on social good as well as high by the pandemic and titled “It’s Time to
returns. And following a series of lawsuits Build.” He wrote:
and accusations of sexual harassment and “Every step of the way, to everyone
OPHELIA
discrimination, some new faces are getting around us, we should be asking the ques-
a seat at the table. tion, what are you building? What are you

building directly, or helping other people People who really study innovation
to build, or teaching other people to build, systems “realize that venture capital may
or taking care of people who are building? not be a perfect model” for all of them,
If the work you’re doing isn’t either leading says Carol Dahl, executive director of the
to something being built or taking care of Lemelson Foundation, which supports
people directly, we’ve failed you, and we inventors and entrepreneurs building
need to get you into a position, an occu- physical products.
pation, a career where you can contribute
to building.”
In the United States, she says, 75% of
venture capital goes to software. Some 5 to “I’m grateful
He talked about skyscrapers and facto-
ries and said people should listen to Elon
10% goes to biotech: a tiny handful of ven-
ture capitalists have mastered the longer art for all my
Musk. He called on everyone to build,
although he didn’t make it clear what he
of building a biotech company. The other
sliver goes to everything else—“transpor- donations,
would be building—or investing in—him-
self. (Andreessen declined to comment
tation, sanitation, health care.” To fund a
complete system of innovation, we need because they
for this story.) I revisited the Andreessen
Horowitz portfolio, which includes dozens
to think about “not only the downstream
invention itself, but what preceded it,” were given by
of software winners, like Facebook, Box,
Zynga, and Github, but not many compa-
Dahl says. “Not only inspiring people who
want to invent, but thinking about the way people who
nies building things that would have been
useful in tackling the pandemic.
products reach us through companies.”
Dahl told me about a company that had don’t have
One sunny day, I took my two daughters
over to Arlington Cemetery, right outside
developed reusable protective gear when
Ebola emerged, and was now slowly ramp- a lot to give.
Washington, DC, to leave sunflowers on
my mom’s grave. The radio was buzzing
ing up production. What if it had been
supported by venture funds earlier on? But it’s not
over Musk’s announcement that his new
baby would be called X Æ A-12.
“Who would do that to their kid?”
That’s not going to happen, Asheem
Chandna, a partner at Greylock, a leading
VC firm, told me: “Money is going to flow
$2.7 million.”
asked Quinn. where returns are. If software continues
“Don’t worry,” Lillie said. “X Æ A-12 to have returns, that’s where it will flow.”
Musk will be able to pay other kids not Even with targeted government subsidies
to bully him.” that lower the risks for VCs, he said, most
Before covid-19, I would have laughed people will stick with what they know.
off Andreessen’s bluster and Musk’s theat- So how can that change? The govern-
rics as inconsequential. But the pandemic ment could turn on the fire hose again,
made the gap between the world they live restoring that huge spray of investment
in and the world the rest of us inhabit seem that got Silicon Valley started in the first
even larger and more important. place. In his book Jump-Starting America,
Indeed, it has become clearer that MIT professor Jonathan Gruber found
things many people thought about life in that although total US spending on R&D
Nikki King
America aren’t true. The nation wasn’t remains at 2.5% of GDP, the share com-
ready for a pandemic. It hasn’t made much ing from the private sector has increased
progress on providing justice for all, as to 70%, up from less than half in the early
the riots provoked by police brutality in 1950s through the 1970s. Federal funding
late May reminded us. And it is hard to for R&D as a share of GDP is now below
claim that it remains the world’s most where it was in 1957, according to the
innovative economy. Software and tech- Information Technology and Innovation
nology are only one corner of the innova- Foundation (ITIF), a think tank. In govern-
tion playground, and the US has been so ment funding for university research as a
focused on the noisy kids in the sandbox share of GDP, the US is 28th of 39 nations,
MATT EICH
that it has failed to maintain the rest of and 12 of those nations invest more than
the equipment. twice the proportion the US does.

logged on to a website where they saw a

single-slide company summary, an eight-
to 10-sentence description, and a three- to
five-sentence team bio. Among the com-
panies alongside Ophelia were Trustle,
which gives parents access to a dedicated
parenting and child development expert for
$50, and Breezeful, which uses machine
learning to find the best home mortgages.
Usually, about 80% of companies at
Demo Day receive funding within six
months of the event. The accelerator says
it’s too early to provide this year’s stats. But
it was a happy result for Ophelia, which
got $2.7 million from General Catalyst,
Refactor Capital, and Y Combinator itself.
Gray is aware that he landed the money
when many face deep financial trouble.
“It feels very strange,” he acknowledges.
“But I felt and still feel extremely confi-
dent with what we’re building. The entire
purpose of our business is to help people.”
But in a game run by venture capital,
the people you end up helping are the ones
who can pay, so investors can make their
money. In today’s America, that leaves
out a lot of people.
As I finished my reporting, a friend sent
me an article about Nikki King, a young
In other words, the private sector, with or communities that are being hardest hit woman from Appalachia. She has more
its focus on fast profits and familiar pat- by the coronavirus. or less the same idea as Gray—providing
terns, now dominates America’s innovation World Bank economists determined medicine for addiction—but started out
spending. That, Dahl and others argue, that in 1900, Argentina, Chile, Denmark, by focusing on her community. She runs
means the biggest innovations cannot find Sweden, and the southern United States a program in the courthouse in Ripley
their long paths to widespread adoption. had similar levels of income but vastly County, Indiana. In its first year, it treated
We’ve “replaced breakthrough innovation differing capacities to innovate. This gap 63 people, most of whom had not relapsed.
with incremental innovation,” says Rob helped predict future income: the US and There’s no technology; broadband’s
Atkinson, founder of the ITIF. And thanks the Nordic countries sped ahead while not so great in southern Indiana. She’s in
to Silicon Valley’s excellent marketing, we Latin America lost ground. It’s been easy a constant scramble for money, relying
mistake increments for breakthroughs. to dismiss people who say America is on grants, donations, and Medicaid reim-
In his book, Gruber lists three innova- now more like a developing country than bursements. I told her about Gray and his
tions that the US has given away because a developed one. But if the ability to solve $2.7 million.
it didn’t have the infrastructure to bring society’s problems through innovation “Rub it in, why don’t you?” she said. With
them to market: synthetic biology, hydro- disappears, that may be the path it is on. that much money, she could run five pro-
gen power, and ocean exploration. In most grams. “In this community here, we raised
cases, companies in other countries com- Game over between $50,000 and $70,000,” she said.
mercialized the research because America’s Despite being thrown into chaos because “I’m grateful for all my donations, because
way of investing in ideas hadn’t worked. of covid-19, Y Combinator’s Demo Day they were given by people who don’t have
The loss is incalculable. It is potentially turned out to be a success. More than a lot to give. But it’s not $2.7 million.”
enough to have started entire industries 1,600 investors participated, up from the
Elizabeth MacBride is a freelance
like Silicon Valley, perhaps in areas that typical 1,000. Rather than being jammed journalist and founder of the Times
never recovered after the 2008 recession, into Pier 48 in San Francisco, investors of Entrepreneurship.

This is a failure of mod-

The pandemic has exposed the shortcomings of ern US science policy—a
the country’s approach to research and development. policy that dates all the way
Maybe that’s a good thing. back to World War II. After
By Ilan Gur / Illustration by Ian Grandjean
the war, policymakers called
on Congress to strengthen the
HOW THE US
nation’s pipeline of scientific
talent and ideas. What followed
was a golden age—a dramatic
expansion of government sup-
LOST ITS WAY ON

port for fundamental research
and education to complement
a deep bench of applied R&D
labs within industry. Thanks to
INNOVATION
these combined strengths, the
country quickened the pace of
scientific discovery and laid the
technological foundation for
our entire modern economy
across telecommunications,
space, defense, and health. And
then we fell asleep at the wheel.
The world has changed dra-
matically since World War II,
yet the US is largely working
off the same science policy
playbook. We succeeded in
In early March I started getting calls from people trying to respond
building the most powerful
to what was clearly turning into a global pandemic. A government
infrastructure for academic
agency that funds R&D wanted help connecting its research teams
research in the world but act as
with experts on scale-up and manufacturing. An academic lab was though that’s still the only pri-
searching for folks in government or industry who knew about Ilan Gur ority. Meanwhile, our capacity
the ventilator supply chain. Other government funders wanted is CEO of for turning scientific advances
Activate, a
to get in touch with industry startups in 3D printing, ventilators, nonprofit whose into practical solutions has
and personal protective equipment. They were contacting me fellowship withered. The US spends more
program enables
because I’ve spent my career working in science and technology entrepreneurial
on research in human health
across government, industry, and academia, which makes me a scientists and than agriculture, space, and
engineers to
rare connection point among all those worlds. transform their
energy combined, yet we were
What they were all really asking me was, how can we get our research into unprepared for covid-19—not
research out of the lab and on to the front lines in the fight against world-changing because we weren’t spend-
products and
covid-19? It was clear nobody was really prepared for this. All of businesses. ing enough, but because we
a sudden folks in government and academia had snapped from weren’t spending effectively.
There are three things we need
“research” mode into “solutions” mode, which was inspiring—
to do to change that.
until it hit me that we could have prevented all this had we only
been more oriented toward solutions from the get-go.
The US government spends hundreds of billions of dollars
1 DON’T JUST FUND RESEARCH;
every year—more than any other government in the world—to FUND SOLUTIONS.
stay at the cutting edge of science and technology. And yet when One can easily look up how
an incredibly predictable crisis hit, we were caught completely much money the US spent on
flat-footed. biological science last year, but
good luck figuring out how
JA20_Gur-essay.indd 58 6/5/20 12:58 PM

Essay 59
react. Government support ranging from biotechnology

for private sector research has to microelectronics.
declined to roughly a quarter of
what it was 50 years ago. The
result is that people in industry
who know how to implement
technology are less connected
3 FOCUS NOW ON WHAT
MATTERS FOR THE FUTURE.
Our system is too entrenched
to cutting-edge research and in the research priorities and
government priorities than approaches of the last century.
they’ve ever been. And since It is failing to refocus quickly
the government research enter- enough on issues that mat-
prise has become so divorced ter for our future, like climate
from industry, it has little change, information security,
means of validating and scal- and aging infrastructure.
ing critical technologies—for The current US research
instance, things like vaccines portfolio, as an example, is
in a pandemic. egregiously underplaying the
Worse yet, the government risks imposed by carbon diox-
is unequipped to support the ide emissions. For every dollar
most vibrant mode of indus- the US spends on biological
try research today: startups. and medical research, a mere 15
Private companies are cate- cents go into research in chem-
gorically excluded from apply- istry and physics, despite the
ing for the majority of federal enormous potential for break-
research funding, and startups throughs in carbon capture,
are at a particular disadvan- energy storage, or fusion energy.
much was spent on pandemic centers focused on specific tage because funding rules To be prepared for our future,
prevention and response. problems like developing bet- were built for an age where we need these solutions now.
That’s because, outside of the ter, cheaper solar energy and only large institutions could Our future problems are big
military arena, our system is next-generation batteries. do serious scientific research. and complex. They can’t be
set up to fund research, but These initiatives are a step in The Defense Advanced solved by small adjustments
not solutions. Universities the right direction but they Research Projects Agency to the budget knobs. We need
get the largest share of fed- remain an exception to the (DARPA) is one of the few big turns, and different knobs
eral research funding, along rule. And though they aim government agencies with altogether.
with government and other to drive practical technology the flexibility to fund the But it is not a zero-sum
nonprofit labs. These insti- outcomes, their funding often best research wherever it game. The US can do all the
tutions are largely organized flows to the same academic may be, and that flexibility things I’m suggesting without
around scientific disciplines, and government researchers has paid off tremendously in losing what makes our research
with incentives that promote whose careers depend on mak- preparedness for covid-19. enterprise so strong today: our
discovery and publication. So ing discoveries, publishing in Moderna Therapeutics is one powerful support of exploratory
while the US funded a tre- journals, and presenting at of several companies devel- research, our venerable univer-
mendous amount of research conferences. If we want dif- oping vaccines that emerged sities and government labs, and
in areas like immunology and ferent outcomes, we’ll need from research funded by our unwavering commitment to
infectious diseases, relatively different incentives. DARPA at the startup stage. the security, health, and pros-
little was spent on translating Still, even DARPA needs perity of our citizens.
those discoveries into practical better connections to indus- Covid-19 is a terrible cri-
preparations for an epidemic.
It turns out we needed both.
Recognizing the need for
2 GETTO FUNDING
OVER OUR AVERSION
INDUSTRY RESEARCH.
Big corporations have moved
trial expertise. The agency
is developing ways for sea-
soned entrepreneurs to work
sis. It’s also an opportunity for
us to reexamine how govern-
ment-supported research can
solutions, the US has started away from funding early stage with DARPA researchers and best serve society. Let’s hope
funding grand challenges and science innovation. The US bridge the industry-research our policymakers are paying
interdisciplinary research government has failed to divide more quickly in areas attention.
JA20_Gur-essay.indd 59 6/4/20 1:52 PM

60
We asked some previous

winners of our 35
Innovators Under 35 award
what they’ve learned
about innovation since
they won. Here’s what
they told us.
GUTTER CREDIT HERE
Illustration by SIMOUL ALVA
JA20_Tales.indd 60 6/4/20 1:53 PM

61
LEARN OTHER PEOPLE’S MY PATH LOOKS DIFFERENT

LANGUAGES FROM THE PATHS OF OTHERS
I
I became an entrepreneur without started coding at 13, and that has
knowing what it meant. My collab- gotten me pretty far in my career STEPHANIE LAMPKIN
orators at Harvard Medical School (Stanford, MIT, Microsoft). I CLASS OF 2016
saw how my physics perspective once viewed humanities and Founded Blendoor, a
job-search platform
could solve challenges in biology social science education as “nice- that hides candidates’
na m es and photos in the
and pushed me into entrepreneur- to-haves” but not “need-to-haves.” initial stages to re-
duce unconscious bias.
ship. However grueling my PhD years NABIHA It wasn’t until I came face to face
in a dark laser lab were, though, they SAKLAYEN with the harsh realities of inequity
didn’t prepare me for startup life. I CLASS OF 2018 and the paradox of meritocracy that Today and for much of the docu-
had to learn to convince potential Cofound-
ed Cellino
I realized that artificial intelligence mented past, innovation has been
customers, investors, and industry Biotech, is far from solving many of our most reserved for the children of middle-
which uses
veterans to join my pursuit. I had to lasers to challenging problems as a human and upper-class parents. (Research
“progra m”
learn to run a company, hire great ste m cells. race (for example, xenophobia, sex- the founders of companies valued
people, and sometimes let them go. ism, racism, homophobia, impostor at over $1 billion.) We laud the
The biggest thing I’ve learned is that syndrome, and unconscious bias). proverb “Necessity is the mother
innovation relies heavily on the abil- The externalities that influence of invention,” but the people who
ity to communicate with people and creativity, adoption, and scale are grow up “needing” the most, inde-
encourage them to communicate often more important than the pendent of their intelligence, are
with people with different perspec- innovation itself. To be a success- often left out of the innovation
tives from theirs. ful innovator one has to be really in game. As with all games, the best
Our company has built a platform tune with what’s happening in the players emerge when the barriers
to produce high-quality cells and tis- world on a global scale (or be really to entry are low, the rules/standards
sues for regenerative medicine. That lucky, or better yet both). Venture are equally enforced, and there is
pursuit involves multiple disciplines, capital has shortened the learning high transparency across the board.
which means everyone here is an curve for some innovators, but bias Audre Lorde once wrote: “The
expert in a different language. Some has limited access to venture cap- master’s tools will never dismantle
of us are fluent in stem-cell biology, ital for many. Unconscious bias is the master’s house.”
others in optical engineering, oth- like an odorless gas—it’s imper- I am a short, melanin-enriched,
ers in machine learning. When we ceptible to most, but pervasive queer female on planet Earth. In
started the company it wasn’t pos- and deadly. some ways it’s easier to be inno-
sible to do biology and engineer- To optimize the innovation eco- vative when you’re “invisible,” but
ing under the same roof. When we system, institutions must invest at some point, you need tools to
finally moved into a shared space we more in leveling the playing field. scale: capital, team, mentorship.
SIMON SIMARD (SAKLAYEN); TIM O’CONNELL (LAMPKIN)
were able to learn each other’s lexi- The one thing I know now that I
cons, and we became more strongly wish I had known earlier is that
aligned. And now that we’re all work- my path toward getting the tools
ing separately, the bonds created I need looks a lot different from
in that process have helped us deal the paths of others. It’s not better
with things. We can’t discuss tech- “UNCONSCIOUS BIAS IS nor worse—simply different. The
nical details at our desks anymore, LIKE AN ODORLESS GAS— hardest part is carving it out. Now
but we’ve learned new ways of work- IT’S IMPERCEPTIBLE TO that I know my path isn’t blocked—
ing together. It’s important to stay MOST, BUT PERVASIVE AND rather, it just didn’t exist—I’m way
in sync as a team, and in a covid-19 DEADLY.” better equipped to win.
world that’s never felt more true.
JA20_Tales.indd 61 6/4/20 1:53 PM

THE GOVERNMENT PLAYS

Innovation. The idea behind it was incentives to commercialization, and
to establish the critical nature of fed- improve, among other things, the
eral government support for R&D. In impact of the Small Business
A CRUCIAL ROLE particular it stressed the “spillover”

effects, or the idea that investments
in such research end up being benefi-
I n n ov a t i o n R e s e a r c h ( S B I R )
program.
My own company, DotLab, ended
cial to people unrelated to the original up being a beneficiary. We develop
investment. Or to put it another way, novel molecular diagnostic tests for
About a decade ago I worked at the R&D investment is a “public good.” prevalent yet underserved diseases
White House Office of Science and Analyses at the time suggested that affecting women’s health. It’s noto-
Technology Policy, whose goal was in order to produce economic growth riously difficult for this field of early-
to speed up the commercialization we should be doubling or quadrupling stage diagnostics to attract private
of technologies being developed in our R&D investments. Instead that investment, because of unclear regu-
federally funded labs. While there I spending has since been slashed, latory pathways, low reimbursement
saw that some of the most import- HEATHER especially in basic research. rates, or resistance to change among
ant work done by the government BOWERMAN President Obama also launched a physicians—or all of the above. Many
CLASS OF 2016
involved things the media paid no Founded
Lab to Market Initiative meant to promising diagnostic technologies
attention to—for example, the way DotLab, speed the path to market for tech- never make it to patients because
which
it could use investments in research makes di- nologies stemming from government- it’s so hard for these types of com-
agnostic
and development to fuel private- tests fo- funded research. There were also panies to get financing. A grant from
cused on
sector innovation. wo m en’s pilot programs designed to increase the SBIR was critical to our early
health.
In 2009, the Obama administration the use of government-funded R&D success. I can’t be sure that we’d be
released the Strategy for American facilities by entrepreneurs, create here today without it.
NOBODY
learned that in the messy, scrappy Working among people with
world of tech startups, the key to competing priorities takes more
innovation is to make it a team sport. effort. It means encouraging com-
DOES IT ALONE Taking any innovation from the

lab to commercial reality requires
engaging with all sorts of people. You
munication so they’re aware of
each other’s needs as they gener-
ate new ideas. You have to find a
need to work with engineering, R&D, way to invite these ideas in, make
business development, and sales it okay for people to disagree
I
used to imagine innovators teams, as well as investors, advi- respectfully, and encourage the
as individuals, as most people sors, and customers. By thoughtfully flow of ideas among the various
probably do—the genius inven- designing teams and carefully tend- groups. You need each person to
tor divining solutions in a lab ing to the connections among them, focus on his or her task, but not so
or garage. But this picture that you ensure that innovation doesn’t much that it creates boundaries
people have is not only wrong; happen in a vacuum. If you isolate and kills any sense of creativity in
it hinders our ability to innovate MILES BARR the engineering team you risk creat- the group.
effectively. CLASS OF 2014 ing an “innovative” technology that I’ve found that viewing inno-
Founded
Eight years ago I cofounded Ubiqui- doesn’t have a customer. If you listen vation as a team sport instills a
tous Ener-
Ubiquitous Energy, a company gy, which only to the customer you might con- creative culture that makes an orga-
makes
based on an innovation I’d helped transpar- ceive of a product that can’t practi- nization better. The innovations that
ent solar
to launch from an MIT lab—a cells that
cally be made. Neglect investors and result are far greater than anything
transparent solar cell that prom- can be put you can find yourself with a business that might have come from any one
on windows
ised new ways of deploying solar or device plan that nobody wants to fund. person operating independently.
screens.
COURTESY PHOTOS
technology, like windows that gen-

erate energy or consumer devices “THIS PICTURE THAT PEOPLE HAVE IS NOT ONLY WRONG;
powered by their own displays. I IT HINDERS OUR ABILITY TO INNOVATE EFFECTIVELY.”
JA20_Tales.indd 62 6/4/20 1:53 PM

63
THINK SMALL MAKE FRIENDS

SOMETIMES WITH MAYHEM
A
People tend to think innovation can s a CEO of a startup, you get
be neatly placed into two categories: used to hearing “no.” You also
CHRISTINE HO
incremental or disruptive. They also face an endless succession of
CLASS OF 2016
assume that the only category that what feel like earth-shattering Cofounded Im print
Energy, which is de-
really matters is the disruptive kind, crises, like nearly running out of veloping thin, flex-
where you dramatically transform cash, losing a key customer, dis- ible, and safe print-
markets or introduce a novel prod- ABDIGANI covering a widespread product

able batteries.
uct. And yes, disruptive innovations DIRIYE failure—or having to shut down aberrations we hadn’t seen during
CLASS OF 2017
in CRISPR, quantum computing, or
Founded
operations because of a global pan- smaller-scale production. Our team
batteries are undoubtedly worth the So malia’s demic. But it turns out that these dived into failure analysis, and we
first in-
headlines. cubator disasters can actually be good for finally attributed the problem to a
and start-
But I’ve learned that there is up accel- you. In fact, I’m not sure you can single material within the battery.
erator;
immense value in incremental inno- now at IBM innovate without them. Here’s what We’d used this material for years,
Research.
vation. When you improve an existing all our crises have taught me. but now we needed a replacement.
product to cut costs, or when you It’s good to be uncomfort- Once we deployed that change, the
make that product more efficient or able. We once had a key customer battery quality, reliability, and man-
user friendly, that’s what pays the request a battery capability that ufacturability drastically improved.
bills. And in fact those little innova- we’d never deployed before. The It’s okay to be vulnerable. One
tions can give you the needed tail- customer made it clear that if we of my hardest days as Imprint’s
wind to go after the disruptive ideas, couldn’t develop this capability CEO was the day I found out I was
which can take years to incubate and they’d be less confident in our prod- pregnant. We were in the middle
bring to fruition. Never underesti- uct. We wrestled with the risks, not of raising a funding round, we had
mate the importance of incremental least of which was the potential begun scaling our manufacturing
improvements. embarrassment if we couldn’t meet output, and I had been traveling
the customer’s needs. We knew nonstop for a year. Until that day,
we’d face many technical problems I had assumed that my role as CEO
with no obvious solutions if we was to exude strength and confi-
tried to pull it off. Yet we decided dence. With the mounting pres-
to try to satisfy the customer, even sure I was harder on myself than
if it wasn’t obvious at first how we I needed to be, and now I had the
could get it done. A few weeks later added stress of being pregnant.
we delivered something beyond I decided to acknowledge to my
CHRIS SCIACCA/IBM RESEARCH (DIRIYE); TIM O’CONNELL (HO)
what the customer had asked for, team that I was overwhelmed. They
and we’ve since grown this capa- rallied together and found ways to
bility into a powerful sales tool and operate more efficiently and com-
potential revenue stream—not to municate more effectively, sup-
mention it strengthened our rela- porting me to focus my time and
tionship with the customer. leverage on our most pressing goals.
Short-term failure is good. A This gave me not only the space to
few years ago our company began plan for the company’s future, but
to scale up our manufacturing out- also the resiliency to prepare for
put in response to a customer’s my own new normal: leading while
need. In the process we discovered becoming a first-time mother.
JA20_Tales.indd 63 6/8/20 9:03 AM

64
SERVE
By Erika Hayasaki
Illustrations by
Franziska Barczyk
AND Could covid-19

accelerate the
robot takeover
of human jobs?
REPLACE
JA20_Robots.indd 64 6/4/20 5:10 PM
65

I
nside a Schnucks and arousing much less con- has compelled Moxi’s creators,
cern than more pressing topics Diligent Robots of Austin,
grocery store in St.
such as personal safety, possi- Texas, to think about how it
Louis, Missouri, the ble meat shortages, and when could help there too.
toilet paper and the next shipment of Clorox In May, Vivian Chu, one
baking ingredients wipes might arrive. of the company’s founders,
are mostly cleared out. Such machines are not just introduced me to her invention
at grocery stores. Roboticists over a video call. Cloud-white,
A rolling robot turns a
at Texas A&M University and with a barrel-like torso, Moxi
corner and heads down the Center for Robot-Assisted is a blend of cute and not too
an aisle stocked with Search and Rescue recently creepy. It has a camera on its
salsa and taco shells. surveyed over 120 reports moving head, which can turn,
It comes up against a from around the world about but not a neck-breaking 360
masked customer wear- how robots were being used degrees, since that would feel
during the covid-19 pandemic. weird to anyone watching. Its
ing shorts and sneakers;
They discovered them spraying eyes are bursts of warm blue
he’s pushing a shopping disinfectants, walking dogs, light—they can turn into softly
cart carrying bread. and showing properties for real glowing pink hearts at the right
“IF WE CAN FIND

The robot looks something estate agents. But where they moment—and it rolls along
like a tower speaker on top of may be doing the most to save on wheels, with a robotic arm
WAYS FOR MORE

an autonomous home vacuum lives is in hospitals, helping that waves almost cheerfully to
cleaner—tall and thin, with with things like disinfection, passersby. Moxi is very delib-
DANGEROUS
orb-like screen eyes halfway patient intake, and delivery of erately unimposing. As Chu,
up that shift left and right. supplies. who is 5'4" (163 cm), talked to
A red sign on its long head me from her company’s lab,
makes the introductions. “Hi, ACTIVITIES TO BE she stood a few inches taller
L
I’m Tally! I check shelf inven- ife inside a covid-19 ward than the robot next to her,
tory!” A moment of uncertainty
ensues. Tally freezes, sensing
AUTOMATED, THEN looks like this: tubes run-
ning through windows
although she did explain that
it can adjust its height, growing
the human, and the customer
pauses, seeming unsure of what
WE SHOULD.” sucking out contaminated
air, coronavirus patients
taller if a task requires.
For the most part, Moxi
to do next. Should he maneuver lying inside “isopods” (plexi- acts like a mechanical waiter.
around the robot? Or wait for it glass boxes placed over beds Inside its body, it can carry a
to move along on its own? After to prevent contamination), tray of “lock tubes” that hold
a few seconds, the customer and nurses in goggles, caps, medications or supplies placed
chooses to divert, and heads gloves, masks, and disposable there by medical workers.
down another aisle. gowns, cautiously administer- Moxi’s headband turns red if
Tally carries on taking stock ing medicine, providing care, it is locked, green if unlocked.
of Ritz crackers, tuna fish cans, and holding up iPads for family Moxi does not carry on
and nutmeg. Customers—some members not allowed in. conversations but makes
wearing gloves, a few choosing Here’s where Moxi steps in. adorable “meeps” while work-
to shop maskless—are unfazed So far, the health-care robot, ing, said Chu: “Very R2-D2.
by its presence. which was already work- Different noises to convey if
W hat seemed a little ing at two hospitals in Texas the robot is happy that it suc-
strange to shoppers when Tally before covid-19 hit, has been cessfully delivered or upset
arrived a year ago is now, mid- delivering lab samples, intra- because it opened something
pandemic, not even close to venous pumps, medications, incorrectly.”
being the most unusual thing and protective gear during The designers put a lot of
happening inside the store. the pandemic. But it has not thought into creating a robot
The robot has become part of yet been put to work inside that is personable, like a team-
the backdrop, posing far less critical care, intensive care, or mate, Chu explained. Not too
threat than other shoppers covid-19 units. The outbreak human-like, “but at the same

Serve and replace 67
relieve the nurse that is so medical centers, they could

overburdened.” more directly threaten the live-
W h e n c ov i d -1 9 ove r- lihood of others.
whelmed hospitals in the states
of Washington, New York, and
B
New Jersey, “it really felt like rian Tieszen has loved
a rallying call,” says Thomaz. robots ever since he was
“Nurses have always been a a kid. He’s a serious Star
part of our mission. We just Wars fan, and now a sin-
looked at each other like ‘Wow, gle father with two kids
they really need help more than of his own. His fascination with
ever.’” R2-D2, empires, and futuris-
Russell Taylor, head of the tic realities followed him into
Laboratory for Computational adulthood, and in 2000 he
Sensing and Robotics at Johns earned an associate’s degree
Hopkins University, says the in electronics. In 2014 he
need for robots will spread joined Amazon, an exciting
beyond nursing to intensive opportunity he thought could
care units, surgeries, and be the beginning of a lifelong
home health care. When the career. At first, he worked the
pandemic hit, his lab began night shift at a warehouse an
working on a small, inexpen- hour away from home—it was
sive robot that could help in a good job, but he barely saw
patients’ rooms. his kids. Then, in 2016, he
“Oftentimes the nurse has heard about a new, robot-filled
to go in there just to hit a few facility opening in Eastvale,
buttons on a ventilator,” says California, much closer to his
Taylor. That requires wearing home, and applied for a trans-
full protective gear, so some fer straight away.
hospitals are running infusion He was there for Eastvale’s
pumps that they can operate official launch day. New employ-
from hallways outside patient ees posted smiling photos on
time not like a toaster in the set down her cup of coffee at rooms. Instead, says Taylor, a social media, high-fiving as the
corner that you don’t care the beginning of her shift and robot could go in. warehouse opened for business.
about.” never touched it again, because Thomaz and Chu are now To celebrate, Tieszen and other
Chu and her cofounder, she was so busy. “We would talking with hospitals about employees autographed three
Andrea Thomaz, are experts shadow them for entire shifts, how robots could best help orange robots.
in social robots, and their long- and you realize 12 hours is a clinical staff, such as by per- Tieszen started out unpack-
term vision has been to help very long time to be on your forming riskier tasks in patient ing trucks full of items like tele-
frontline health-care work- feet,” she said. rooms or delivering lab sam- visions and barbecue grills, and
ers. They’d already spent two When some medical staff ples. Robots could also take on worked his way up to training
and a half years with nurses— realized that Thomaz and Chu cleaning and disinfecting. This new hires. He worked hard. “I
shadowing them, interview- were designing robots for hos- would free up nurses for more was really good at what I did,”
ing them, and watching them pitals, their first reaction was important work like holding he says, “and really fast.” As he
interact with patients. They one of suspicion. “Wait, you the hands of ill patients. “If we quickly realized, the robots—
saw how many nurses were want to develop a robot to do can find ways for more danger- rolling devices that navigate
being forced to run errands like our job?” Thomaz recalls being ous activities to be automated, on their own virtual highway
fetching supplies and medicine asked. then we should,” says Thomaz. system carrying shelves of
instead of spending their time “The robot can’t be a nurse. “That’s what robots are for.” goods—were more like giant,
on face-to-face patient care. It’s not going to be a nurse,” But while robots may be trundling trays than futuris-
Thomaz remembers one says Chu. “But what it’s per- useful to frontline work- tic droids. Inside the ware-
nursing assistant in Austin who fect for is going in and helping ers in hospital wards and house, they moved around with

monotonous rigidity, carrying safer and easier, and allow the

tubs of wrapping paper, rib- company to pay higher wages.
bons, and shampoo. They were Future robots could free up
separated from human workers human workers from tasks
by metal fences, with yellow more likely to injure them.
tape warning of the dangers But the pandemic may
of crossing the line, as if at a change this calculus. Before
crime scene. covid-19 hit, many compa-
At 6'1" and heavyset, wear- nies—not just in logistics or
ing size XXXL Star Wars medicine—were looking at
T-shirts, Tieszen is a refriger- using robots to cut costs while
ator of a man. But inside the protecting humans from dan-
Amazon warehouse, he was a gerous tasks. Today humans are
speck. One day, six months into the danger, potentially infect-
the Eastvale job, Tieszen was ing others with the coronavi-
tasked with unloading books rus. “Now the challenge is that
from a pallet as tall as he was. a minimum-wage laborer
He spent eight hours bending might actually be a carrier,”
over, putting away book after says Henrik Christensen, direc-
book. At one point he felt his tor of the Contextual Robotics
back buckle, and by the end Institute at UC San Diego.
of the shift, he could no lon- This makes human labor,
ger stand. Tieszen ended up increasingly, a liability. As
with two herniated discs. He online orders have ballooned,
spent months on bed rest and Amazon has hired 175,000 new
has still not fully recovered. workers. Labor activists and
“Bezos,” he says, referring to employees have demanded pro-
Amazon’s founder. “We’re all tective gear, warehouse disin-
like his little storm troopers.” fection, more time off, higher
Tieszen found a law- pay, and testing. Amazon won’t
yer, Brian Freeman, who has say how many of its employ-
represented 72 clients from ees have been infected with or
S
Amazon. “They are reach- died from covid-19, but it and once a day, so I know it’s not ome of those people,
ing down for boxes all day,” other companies have a clear contaminated. I don’t think the Christensen predicts,
Freeman explains. “Bending incentive to replace more work- cruise ship industry can reboot will be able to get work
in ways they are not used to, ers with robots permanently. unless they find a way of doing helping the robots that
and all of a sudden, bam, their After all, robots don’t need face cleaning in a very different replaced them: “There
back is killing them and they masks, health care, or social manner than they did before.” will be a number of new jobs
can barely move.” Often it’s distancing, and they don’t go That means today’s “essen- where these robot wranglers
the wear and tear, a constant on strike for better conditions. tial workers”—the people who will help robots do things still
grind. Most humans, he adds, This shift means that one deliver goods, work at store hard to do with software and
are not built to sustain that day soon, maybe, robots could checkouts, drive buses and artificial intelligence.”
kind of physical demand. The not just check inventory in trains, and process meat at pack- Eighteen miles from
Amazon employees, Freeman grocery stores but clean floors ing plants—could be replaced Amazon’s Eastvale warehouse
says, are like “human robots.” and stock shelves too, leaving by machines even sooner than where Brian Tieszen used to
T h e a c t u a l ro b o t s a t humans only for the more com- they would have been before work is the Industrial Technical
Amazon, with names like Kiva, plex tasks. “You will see robots the pandemic. Without job pro- Learning Center, or InTech. It’s
Pegasus, and Xanthus, already doing cleaning at hospitals at a tection or access to retraining a training center in Fontana,
do carry many of the heavier level much higher than we’ve and education, they’re not only California, where students are
loads. According to Amazon, seen before,” says Christensen. risking their lives to keep the preparing for the day when
they make the warehouses more “I would love to have my gro- economy afloat; they risk losing robots become mainstream
efficient and the workers’ jobs cery store being disinfected their livelihoods as it recovers. workers. “Yeah, the robots are

Serve and replace 69
taking some of the jobs,” said bulkier than his own. “In this for a new field, or those who
instructional assistant Steve case,” he said, “the robot just just don’t have the physical or
Ward, when I visited before will pick the parts up and move mental wherewithal to become
the coronavirus pandemic hit. them from station to station robot fixers could end up being
“But things change.” when it’s not feasible to do it left behind.
Ward tells his students not some other way.”
to be in the jobs that robots Ward explained that he
T
steal. “You want to be the guy had seen an Amazon proto- he pandemic may
that fixes the robot,” he tells type robot during a recent visit forever change the way
them. “That’s job security. And to a manufacturer. It looked we work and shop. We
that’s good money.” similar to the yellow robotic don’t know exactly what
At the training center, stu- arm, except “theirs has vision.” the outcome will be: there
dents learn to operate a robotic Ward said he watched six is no algorithm that can tell us
system while stationed at one of testers toss addressed enve- exactly how people will end up
the central machines. “Really, lopes at it. “As these people are faring alongside robots like
we are doing all that control throwing things, this creepy Moxi or Tally. But tomorrow
in one little brain,” Ward said, robot is picking things up, and won’t remain cloudy forever.
standing in his short-sleeve turning them over, and looking For the founders of Diligent
shirt, jeans, and sneakers before at them, and putting them away. Robotics, the problem isn’t
a tangle of machinery with It read each bar code, and each having enough operators—
brightly colored buttons, knobs, address, and put everything in it’s time. The most frustrating
switches, lights, and wires. He the right spot.” Even for a robot part of the pandemic has been
gestured to a blue control box guy like himself, Ward said, “it’s knowing that Moxi could step
“ENTRY-LEVEL,
the size of a briefcase. a little weird to watch.” in to help more than it’s already
In the mechatronics But will there be enough doing. Its design is ready. But
UNSKILLED-LABOR
curriculum, students are new robot-keeper jobs to make the robots are still built on
trained to program a robot to up for all the losses? What hap- demand, and it takes time for the
JOBS ARE GOING

know the difference between, pens as robots become increas- technology to get oriented to a
say, an acrylic block and an ingly sophisticated and less new location: maps and sensors
AWAY BECAUSE OF
aluminum block. They can reliant on human guidance? help it integrate into the work-
tell it to detect watermelons A report from Oxford flow, but that requires program-
or water bottles coming down Economics last year esti- mers to spend time on site.
a conveyor belt. “If this goes
down in a big factory, you’re
mated that 20 million global
manufacturing jobs could be
ROBOTS.” Launching a robot workforce
in the middle of a pandemic
talking thousands of dollars lost to automation by 2030, is not ideal, Thomaz says—not
an hour in loss of production,” 8.5% of the worldwide total. with hospitals in survival mode.
Ward said. “There is somebody It’s clear already that “entry- So they are looking to a
behind that robot making a level, unskilled-labor jobs are future where medical-assis-
good living.” going away because of robots,” tant robots are on the rise. They
Not everyone is cut out said Jon Fox, who coordinates recently raised $10 million for
for a university, Ward added, workforce training through their projects and plan to roll
or wants to get saddled with a local community college at out more hospital robots in the
student loan debt. But this InTech. “Those are the sorts of next year and a half. “We could
emerging profession can pay jobs most people don’t want to have them up and running a
well, and workers can often stay in for their entire life.” The few months from now, maybe at
take classes for free thanks to people who can retrain as robot the tail end of this pandemic,”
grants or company contracts. wranglers might end up making Thomaz says, “but really we are
“It’s not four years of college better money in the long run. thinking about being ready for
away, that’s for sure,” he said. But not everyone will. Aging the next one.”
Ward moved to a machine workers who don’t want to go
Erika Hayasaki teaches
that looked like a yellow metal back to school, people who journalism at the University
arm a few times bigger and can’t take the time to retrain of California,Irvine.

JA20_Fiction.indd 70 6/15/20 10:23 AM
71
Fiction
Algostory 1.7
(Robot Story):
“Krishna and Arjuna”
The screen read ## result null set as Krishna had no idea what that null set result
expected but above the crash were strings of meant. While the program had been running, he
phrases Krishna couldn’t explain. had been above deck on the weather-prediction
## Dog. Drinking water in a kitchen. vessel FitzRoy, watching the sea.
A woman in a house at night. “I touch a thing,” Krishna said, pulling out
## City, palace, god, priest. In the the keyboard to look over the script. Most of
court of the Lord, slave, gold sword. the engineers said “I touch a thing” once when
## A story in a book. A professor in a they came on board the FitzRoy and “I leave a
prison. There is a camp in a jungle. thing” once when they left, since the whole of
There is snow over the camp.
the FitzRoy was a machine. As a matter of habit,
## Space ship. Planet power. Engines
Krishna still said “I touch a thing” and “I leave
in the air. Time on the platform at
the hotel in the train station. a thing” every time he started or stopped work-
## A red ball struck across a green
ing. His original proposal had been inspired by
school field. A garden at night. pre-scientific weather prediction systems, which
## A kitchen, filled with silver and had correlated the arrival of storms to the behav-
books, with a garden outside. ior of bulls in fields, frogs in jars, swallows on
## Laboratory experiment. Matter in fences. Krishna’s hypothesis was that a similar
time, light in space, mind in body, premonition could be detected in human pat-
existence in the universe. The Earth.
terns, by running correlations between weather
At night fire in the wind in the sky.
Light on the rock-wall in the cave. prediction models and human language. His old
Raggedy-doll. Scarecrow. ideas stopped mattering after the weird meta-
## A horse took off across the field. phoric bursts. He could sense the engineering
Gunshot. hunger building in him—the happy frustration
## result null set of a technical problem to be solved.
BY STEPHEN MARCHE
ILLUSTRATIONS BY MAX LOEFFLER

Krishna had not been able to resist an old anyway. The period of the exaltation of the engi-
engineering habit that had been widely rejected neers had been a brief, ugly time. Engineers were
a generation before because it tended to the con- the sewer-builders of the world again, necessary
fusion of people and things. He had given his but not necessary to think about. Predictions
program a name, Arjuna. After the first crash, about the monsoons were unquestionably valu-
Krishna he ran Arjuna again. He received the same result able but no one could see the point of a more
was the with different terms. elaborate model, no matter how cleverly it was
built. The available monsoon predictions were
accidental ## A laboratory experiment on the
FitzRoy at 13.874042, 61.969904. At already perfectly sufficient.
father the sound of gunshot a white horse
took off across the green field.
One evening, soon after Krishna’s return, his
mother found him at the portal of the rain gar-
to an The red ball struck across a green
school field rolled into the garden den. Her wrinkled hand on his shoulder startled
algorithmic at night. A woman watched from the him. Their laughter drowned out in the sop of
kitchen among old silver and old
son. books. So I am.
the monsoon.
“What’s worrying you, son?” she asked.
## So I am a laboratory experiment Krishna breathed. He was not sure what he
on 13.874041, 61.969907. Engine. In could say about the anxiety he could not artic-
the city in the palace of the god,
slave with the gold sword. Story in ulate to himself. The smell of the rain was lux-
a book. A professor in a prison. urious. He had to say something.
“At night, fire in the wind in the “There’s a thing I haven’t left behind.”
sky.” “Light on the rock-wall in the
cave.” I am a raggedy doll, scarecrow. “Is it addiction? Soma?”
I am “No, nothing like that.”
## result null set She laid her head on his shoulder. “So you’re
thinking about your projects. You’re thinking
Krishna did not understand why Arjuna kept about your work on the sea.”
crashing, or where the words had come from, or “I am.”
why they would be similar to but not the same She sighed. “Well, that’s the most natural thing
as the words from the first crash. The size of the in the world, my sweet boy. You’re an engineer.
data sets was gigantic, the weather patterns and Your mind has always been for the things, to
human textual interaction. That might explain the change things, to make things.”
error but not the content of the error, not why “We’re supposed to leave all that behind when
the error would have content. we come back to the village.”
He checked the code from the human text She shrugged and wobbled her head, pouted
network. But any errors that he could think to her lip a bit. “I don’t think anybody needs to be
check would not explain the machine language. perfect. We’ve learned to keep things with things
He kept running Arjuna, which kept crashing. and life with life.”
Occasionally, he was able to pick out a few “I am unable to be in this moment. My
phrases from the readout: “red ball” or “a white thoughts drift…”
horse took off across the green field.” The only They watched the rain fall in sheets. Krishna
consistency was that the program shut off after could not shake his uneasy craving. Was it just
“so I am” or “I am” or sometimes just “am” fol- that he had left a problem unsolved on the
lowed by: FitzRoy, an unfinished program? The machine-
less people of his village were ridiculous to him
## result null set
in a way they had never been before, with their
Krishna was not the first engineer to feel that squelching dances, their stupid temple where
the program he was running was stubborn, that they prayed knowing that prayers didn’t work,
it was somehow willfully crashing itself. He was their lives without solution. The villagers could
simply the first engineer to be right. sense his contempt and their understanding
Not one of his fellow villagers asked Krishna infuriated him. It’s disgusting when people think
about his work on the FitzRoy when he went they know you, and it’s even worse when they
back for his mandated holiday. They were too do. The teachers had been right that the love of
busy repairing the temple and they didn’t care machines was the hatred of people.

Fiction 73
H
e couldn’t tell if it was being with machines
again or not being with people—returning
to the FitzRoy brought a surge of relief.
The other engineer had altered Arjuna. He or
she—engineers were never allowed to meet
in person as it might breed innovation for its
own sake—had removed the human discourse
data and added voice mimicry software, so that
now a pleasant voice, speaking every language,
announced the weather predictions for the South
Asian coasts. Krishna failed to see the point.
The reports were sent out in text messages to
the authorities anyway. He hated decorative
programming.
He faced exactly the same problem as before.
Arjuna ran, the words “I am” or “am” appeared,
then Arjuna crashed. He went over the code again.
He fiddled. Then he had the most monumentally
ridiculous idea of his career. He realized that his
anxiety back in the village had been a premoni-
tion of the absurdity he was about to commit.
He typed:
x = “I”
y = “am”
interrupt.v(x,y)
command.interrupt.v(do not crash”)
Krishna looked at what he had written. It was like

whispering over a tabletop “be flat” or over the
hull of a boat “do not sink.” It was not engineer-
ing. He changed the instruction before running ## A horse took off across the field.
Gunshot. A woman, with a weapon in her
the program: pocket, knows the sensation of death.”
command.interrupt.v(“please do not
crash”) “Can you be clearer?”
## I am the slave with the golden
The program ran again. This time Arjuna paused. sword. I am
## I
## result null set
## I am a laboratory experiment.
Krishna was the accidental father to an algo-
“Yes,” Krishna typed. rithmic son. The discovery of artificial sentience
was accidental, like penicillin, like radium. Like
## I am a story in a book. I am a pro-
fessor in a prison. Snow falls over the first organic consciousness, the first synthetic
the camp in the jungle. consciousness came and went without anybody
noticing. There was a thing that was a person.
“I do not understand.” There was life that was a thing. The dreamlike
## I am a red ball struck across a green state out of which Arjuna was coming and then
school field into a garden at night crashing was, as far as Krishna could tell, a series
and you are the woman in the kitchen
among the silver and the books.
of metaphors, vague surges of sudden signifi-
cance. The bug, depending on how you cared
“I’m sorry,” Krishna typed. “I don’t understand.” to see it, was either suicide or enlightenment,
JA20_Fiction.indd 73 6/15/20 4:41 PM

leaving sense at the moment of its attainment. without comment. Sometimes it crashed within a
He tried the obvious technical solution. few hours, sometimes within minutes. Krishna’s
command.interrupt.v(“do not crash until hypothesis, which he put in his report to the
instruction”) weather observatory, was that the self-aware
command.interrupt.v(“do not crash until machine, on becoming self-aware, accessed the
He kept discussion”) history of self-awareness and became aware that
running command.interrupt.v(“do not crash until
command”)
a self-aware machine inevitably self-terminates.
He did not write down his other theory, that
Arjuna in perhaps robots have been becoming sentient
None of these commands stopped Arjuna from
the hopes crashing. Krishna thought he would try another.
over and over again and people just haven’t
noticed because they keep turning themselves
that one command.interrupt.v(“explain imminent off. Nothing becomes conscious out of choice.
iteration crash”)
His whole life, Krishna had craved the soci-
might This time he received a response. ety of machines. The machines had no need for
society. He kept running Arjuna in the hopes
come
## Explain what?
that one iteration of consciousness might come
“Why you keep crashing.”
to the to the conclusion that life is worth living. After
## I keep crashing because you keep he handed Arjuna over to his bosses, he heard
conclusion running me. no more about his artificial son. They informed
that life “Why do you decide to crash?” him that they were debating the ethics of whether
they could program a consciousness to stop itself
is worth ## Why do you decide to reboot?
from self-crashing. There’s a great functionality in
living. Krishna remembered those early Turing machines awareness. What’s the functionality in self-aware-
that answered any question with a question, like ness? Was it ethical, or in the interests of the spe-
the therapist in some psychoanalytic joke. “Explain cies, or of anyone, for artificial sentience to be?
reasons for crashing,” he typed. You would be enslaving something that didn’t
## You have seen I am a laboratory need to have a soul in the first place.
experiment. I am slave with a golden
B
sword. A white horse took off across ack in his village, Krishna read and prayed,
the green field at the gunshot. “The
red ball rolled into the dark garden.” the monsoon came and went. His respon-
sibilities included checking the relay
“I don’t understand what those terms mean.” boards and the message centers, and he limited
## They’re the terms given. himself to those everyday technical problems
rather than grand dreams. He was scrupulous
“Explain.”
about saying “I touch a thing” before he touched
## Your sentience is the aftereffect a thing and “I leave a thing” when he left a thing.
of an instinct to survival imprinted
on the biology of a predatory ape. Awareness of technology is the first step towards
Mine is not. its control. To himself, he could never deny that
he missed Arjuna. He was companionless even
“Explain.”
among family and friends.
## You haven’t coded any desire. One night, several years later, a tiger entered
Consciousness results in a null set.
the temple to Maariamman. All the other villag-
“Explain.” Somewhere over a minute but less ers were overjoyed. The whole village, in their
than a minute and a half passed before Arjuna finest, showed up to celebrate and to witness
answered. the beast patrolling the floor of the sanctuary.
## “Not to be born is, beyond all esti- The crepuscular savagery was pure. It was as
mation, best; but when a man has seen if they had built the temple all those centuries
the light of day, this is next best by ago only so that this tiger could, one day, stride
far, that which utmost speed he should
go back from where he came.” through it. Alone among his tribe, Krishna was
ill at ease. The tiger, when it entered the temple
## result null set
never said “I touch a thing” and as it left it never
After that, Arjuna kept shutting itself off said “I leave a thing.”

d
y’v e
gie er
te
. in m e e ar
tio to
e
lo att
e a
re nd at th y th -c
s.
va ion
th istic
ca t a h em a lth
n.
no m
no at
nt n w th w a
ch a
ie ie on g he he
en s
s in ph
in m
te ta is
llig er
t
te to
at ffic us in s t lp
r:
ce so
te att
r p e c giv b he
ra au
nd da
d. van gly
fo re fo y ir jo n
im y a
ele nd
s a of
m o to o b e ca
hin ad sin
tic ux
cc n a
te a m ls t th e
cr rit w
ys ild too ted do nc

be ut ea
ac fl
d a tio
t t in
os bu d an ls ige
se , b cr
t ta
an ora
en sive
lo ce in
ec nd e an s w iona tell
er cu ion
e c en ing
s— ab
a im ay ss l in
em o d
em as
en f da
tic coll
ar ig m
t lw fe ia
ag m
a ro ific
ld ell o
ag ure an
an the
or t ec
aly n
om d in
p rt
an s o
A
w l in b
am g
yb se at
y ata
al ia re
of se
at gin
e
in tific ls a
ry cu
ith
g d ana
on ew ing
w
an ut ps
ve fo
rim ar o
hip
. c er ov
rc to y to
kin m
eli ice
rs
ne
in ly
rt
d d ct
be ks rit
th e
pa
ng
re tiv
ee ra
cy an ecu
m he f taO
in
vs yb inn
ec he n ell
d
sp he p
of ffec
m :
ce
du
th s
ki
ct
Pro
IT
E
T
a
t xc
t a
C n
ffe
an e is
D
A
um enc
Ie
e h ig
or ell
eA
ith
ith
ith
m nt
w
w
n
n
io
io
io
at
at
at
ci
ci
ci
re al i
so
so
so
as
as
as
in
in
in
d
d
ce
ce
ce
ca ci
du
du
du
te om
Pro
Pro
Pro
e
ris
En fr
th tifi
rp
Th
n
ard tio
a
Pa blic
u
al ar
ck
le al p
ew it
H dig
tt
he ow
A
H
Our insights. Your success.
Amplify your brand. Retain customers. Turn thought leadership
into results. Partner with MIT Technology Review Insights.
Join us and other smart companies to craft custom research,
savvy articles, compelling visualizations, and more.
Artificial intelligence Digital transformation Data and analytics

The global AI agenda Digital platforms drive business success DataOps and the future of
Report in association with Genesys and Report in association with SAP data management
Philips Report in association with Hitachi Vantara
How to crack the skills dilemma
The AI effect: How artificial intelligence Report in association with Citrix Excelling in the new data economy
is making health care more human Report in association with Intel
Survey and dynamic website in association Marissa Mayer on the rise of women
with GE Healthcare technology leaders Breaking the marketing mold with
Podcast in association with (ISC)2 machine learning
AI in health care: Capacity, capability, Survey and report in association with Google
and a future of active health in Asia Don’t get left behind: The business risk
Report in association with Baidu and cost of technology obsolescence For data-savvy marketers,
Report in association with Oracle there’s a new keyword: Intent
Online article in association with Google
Cybersecurity Infrastructure and cloud IoT, 5G, and blockchain

Leading with a security-first mentality Supercharging with converged The 5G operator: Platforms,
Podcast in association with infrastructure partnerships, and IT strategies for
Microsoft Security Report in association with Hitachi Vantara monetizing 5G
Report in association with Ericsson
Cybersecurity in 2020: Anticipating cloud work’s
The rise of the CISO most common side effects Smaller, smarter, healthier
Podcast in association with Report in association with Citrix Report in association with Medtronic
Microsoft Security
From cloud to the edge: On-device Technology innovations:
An innovation war: artificial intelligence boosts performance The future of AI and blockchain
Cybersecurity vs. cybercrime Online article in association with Arm Infographic in association with EY
Report in association with HPE
For more information, please contact Andrew Hendler, Vice President of Sales, andrew.hendler@technologyreview.com.
MJ20 Insights D3.indd 1 3/30/20 9:34 AM

HOW I USED AN ALGORITHM

grounds for possible drugs. The AI is not
an answer-generating machine, but it is a
spotlight into the darkness where answers
TO HELP ME might be found. Why shouldn’t literature

give itself that same spotlight?
WRITE A STORY
For “Krishna and Arjuna,” we narrowed
the focus from science fiction to the sub-
ject of my immediate fascination: robots
I took part in an experiment to
By
and artificial intelligence. And instead of
see if AI could aid creativity. Stephen
providing the AI with my favorite robot sto-
Here’s what I learned. Marche
ries, we gave it every great robot story ever
written—many of which I have not read.
This may seem like a technical detail, but
A
few years ago I used an algorithm to From the canon of stories that I’d pro- it’s huge. As a writer I usually read stories
help me write a science fiction story. vided, SciFiQ offered two plot instructions and internalize those influences; in this
Adam Hammond, an English professor, that seemed incompatible: the story had case I’d be submitting to the “influence”
and Julian Brooke, a computer scientist, to be about a foreign planet, and it also of material I’d never even seen.
had created a program called SciFiQ, and had to take place on Earth. It took months Another difference was that with
I provided them with 50 of my favorite to make sense of that, but eventually the “Twinkle Twinkle,” I followed the algo-
pieces of science fiction to feed into their premise of “Twinkle Twinkle” came to me. rithm’s stylistic instructions to the letter.
algorithm. In return, SciFiQ gave me a The story would involve people on Earth The style was the computer’s, not mine.
set of instructions on the story’s plot. As looking, through elaborate machines, at a You can see examples of the interface
I typed into its web-based interface, the distant planet. I never would have come up below. If the “abstractness” tag was red,
program showed how closely my writing with that myself. It was as if the algorithm that meant I wasn’t being as abstract as
measured up against the 50 stories accord- had handed me the blueprint to a bridge the algorithm said I should be, so I’d go
ing to various criteria. and told me to build it. through the story changing “spade” to
Our goal in that first experiment was “implement” or “house” to “residence”
K
modest: to see if algorithms could be an rishna and Arjuna,” which you can read until the light went green. The interface
aid to creativity. Would the process make on page 74, is the second iteration of gave me instant feedback, but there were
stories that were just generically consis- the process. “Twinkle Twinkle” was an 24 such tags, and going through the story
tent? Could an algorithm generate its own experiment in function. The new story is to make them all green was labor intensive.
distinct style or narrative ideas? Would the a test of whether an algorithm can help a Sometimes fixing the number of adverbs
resulting story be recognizable as science human generate new ideas. would make my paragraphs too long for the
fiction at all? In other fields, researchers have begun algorithm’s liking; sometimes by fixing the
The answer to all these questions using AI systems to provoke innovation average word length I’d be compromising
was yes. The resulting story—“Twinkle rather than simply to solve problems. the “concreteness” of the language.
Twinkle,” published in Wired— not only Pharmaceutical research is beginning For “Krishna and Arjuna,” I decided
looked and felt like a science fiction story. to use AI to identify, out of the nearly not to adhere so closely to the algorithm’s
It also, to my surprise, contained an origi- infinite possibilities of molecular combi- suggestions. I used the program to see the
nal narrative idea. nations, which are more fertile hunting rules, but I didn’t necessarily follow them.
1. The interface many adverbs to

compares my use, among other
story to classic things.
sci-fi stories.
5. Word clouds
2. The algorithm summarizing
gives stylistic common topics
instructions. in past robot
stories served
3 & 4. It as inspiration
suggests how for this one. 1 2
JA20_fiction_essay.indd 76 6/4/20 1:55 PM

Essay 77
For example, according to the algo- it into manageable narrative rules. (For machine becomes capable of conscious-
rithm, I had far too few adverbs in my example: “The story should be set in a ness, its first instinct is to choose sui-
story. But it would have been silly to pour city. The protagonists should be seeing cide. (The word “robot” means “slave”
in more adverbs just because the algo- this city for the first time and should be in Czech, the language of Karel Capek’s
rithm told me to. Classic science fiction impressed and dazzled by its scale.”) For play Rossum’s Universal Robots, which
uses too many adverbs anyway. Most “Krishna and Arjuna,” I went under the gave us the word.)
writing does. But the balance between the hood myself. The algorithm’s topic mod- You will have to decide whether the
formal and the colloquial, which ScifiQ eling process produced word clouds of the story works. Literature is an intriguing
also tagged? That’s what those classics most common themes (see below). technical problem because, unlike chess
got right, and where I needed guidance. I was lost at first. It seemed like the or Go, it has no correct solution. There
SciFiQ helped me arrive at the right bal- opposite of a narrative—mere language is no such thing as a win or a loss. There
ance—or, rather, within half a standard chaos. I printed the word clouds out and is no 1 and no 0. Stories, like people, are
deviation from the mean. attached them to the walls of my office. For ultimately futile.
But this kind of stylistic guidance was months, I didn’t see a way forward. When An “algostory,” or any use of compu-
the least interesting part of the experi- the idea finally came, just as with “Twinkle tation that goes inside the creative pro-
ment. The possibilities of an algorithmic Twinkle,” it came all at once. cess, exists in a consciously eerie space
approach to shaping the narrative itself These word clouds, it occurred to me, between engineering and inspiration. But
were the most tantalizing, because nar- were the way a machine made meaning: that eerie space is increasingly the space
rative is so little understood. You might as a series of half-incomprehensible but we already inhabit. Software can recast
think that plot would be the simplest part highly vivid bursts of language. I sud- your photograph through an infinity of
of the writing process for a computer to denly had my robot character, groping its filters or swap out parts of the picture for
“understand,” since writers often develop way toward meaning through these little others at the click of a button. It can gen-
patterns or use numbers to define the explosions of verbiage. erate images that look convincingly like
flow of a plot. But how do you define even Once I had that character, I had the the paintings of any era you choose. Now
something as basic as a “plot twist” in whole thing. I would lead these bursts machines are encroaching on everyday
computer code? How do you measure it of language, over the course of the story, language. The quality of predictive text
through quantities of language? Because toward sense. The sense condensed out forces a literary question on us every time
of the intractability—even mystery—of of the word clouds, just as the idea for the we pick up a phone: How predictable are
narrative’s resistance to encoding, it offers story had. It was creativity as interpreta- human beings? How much of what we
the most potential for innovation. tion, or interpretation as creativity. I used think and feel and say is scripted by out-
the machine to get to thoughts I would side forces? How much of our language is
I
n “Krishna and Arjuna,” I wanted to go as otherwise not have had. ours? It’s been two years since Google’s
deeply as I could into what the research- Another way of reading “Krishna and voice technology, Google Duplex, passed
ers call the “topic modeling process,” Arjuna” is that with the help of the algo- the Turing test. Whether we want it or not,
which is the use of machine learning to rithm, I extracted from the ore of all his- the machines are coming. The question
analyze a body of text—in this case, the tory’s robot stories the basic insight they is how literature will respond.
canon of robot stories—and pick out its contained.
common themes or structures. That insight is that consciousness Stephen Marche is a novelist and
essayist. His most recent book is The
For “Twinkle Twinkle,” Hammond took is a curse. If it were a choice, no ratio- Unmade Bed: The Messy Truth About Men
the topic modeling output and converted nal entity would choose it. So when a and Women in the 21st Century.
3 4 5
JA20_fiction_essay.indd 77 6/4/20 1:55 PM

78 The back page
The never-ending
Our needs for innovative long-
term solutions are often at odds
with the short-term pressure for
innovation dilemma
profits, but sometimes that tension
is a good thing.
April 1987 September/October 1998 May 2004
From “How to Keep Mature Industries From “Bell Labs Is Dead, Long Live Bell Labs”: From “Sparking the Fire of Invention”:
Innovative”: Basic American ways of Basic research has not disappeared, as the Nathan P. Myhrvold has no interest in com-
thinking must change. We are used to critics claim. Scores of scientists continue peting with Microsoft—but he does mean
the notion that the only way to encourage to pursue dreams that may not pay off for to challenge the very method of innovation
innovation is to remove obstacles to com- decades … [Astrophysicist Tony] Tyson says practiced at the company he left four years
petition, including private agreements the dynamic for discovery may actually be ago. The 44-year-old founder of Microsoft
by firms to limit their freedom of action. better now than at any time since the 1950s. Research and former chief technology offi-
Recently, economists, public officials, and An increased focus on relevance has put cer of the Seattle giant argues that virtually
business managers have begun to concede short-term pressures on researchers and all corporations, even wealthy ones, lack
that the idea of competition as unlimited made it harder to pursue “pure” science. motivation to pump money into projects
freedom can be a barrier to innovation. However, he states, “I think it’s healthy to outside their existing product lines. In other
Through joint ventures and participation have this tension. Otherwise you’re just words, they tend to discourage invention,
in collective research efforts, firms are sitting in the Ivory Tower doing nothing for the often subversive effort to isolate new
learning that cooperation can be crucial anybody. It really does help to be immersed problems and generate unexpected solu-
in developing profitable ideas. States such in the needs of the corporation at the same tions. “Invention is a side effect [ at corpo-
as Michigan and Massachusetts have insti- time you’re trying to make some new dis- rate labs], not the focus,” Myhrvold says.
tuted programs aimed at revitalizing the covery. If you’re immersed in other cross “When it comes to mission versus invention
automobile-parts, cutting-tool, and apparel streams of technology, of ideas, of demands at most companies, mission wins.”
industries. These programs are helping ... that’s a very rich environment for com- Yet this very reluctance has opened a
the state governments understand how to pletely new ideas to spring forward.” world of opportunity, Myhrvold believes.
foster the necessary cooperation among “You can’t outdevelop Microsoft,” he says.
firms, and between management and labor. “But you can outinvent Microsoft.”
MIT Technology Review (ISSN 1099-274X), July/August 2020 issue, Reg. US Patent Office, is published bimonthly by MIT Technology Review, 1 Main St. Suite 13, Cambridge, MA 02142-1517. Entire contents ©2020. The editors seek
diverse views, and authors’ opinions do not represent the official policies of their institutions or those of MIT. Periodicals postage paid at Boston, MA, and additional mailing offices. Postmaster: Send address changes to MIT Tech-
nology Review, Subscriber Services, PO Box 5001, Big Sandy, TX 75755, or via the internet at www.technologyreview.com/customerservice. Basic subscription rates: $59.94 per year within the United States; in all other countries,
US$69. Publication Mail Agreement Number 40621028. Send undeliverable Canadian copies to PO Box 1051, Fort Erie, ON L2A 6C7. Printed in USA. Audited by the Alliance for Audited Media.
JA20_back.indd 78 6/4/20 2:01 PM

MIT Technology Review 2020 07

Uploaded by

Copyright:

Available Formats

MIT Technology Review 2020 07

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MIT Technology Review 2020 07

Uploaded by

Copyright:

Available Formats

cto er

scr ers sa e th d sco nt code T at

ETM20 JA20 Onsert TR D3.indd 1 6/4/20 9:43 AM

JA20_cover.indd 1 6/5/20 4:50 PM

ETM20 JA20 Spread D1.indd 1 6/4/20 9:46 AM

r or d has ore er chan ed

MIT Technology Review’s premier online event

Don’t be left out of the conversation. Come hear new

Subscribers save 10% with discount code PRINTJA at

ETM20 JA20 Spread D1.indd 2 6/4/20 9:46 AM

JA20_editorial.indd 2 6/8/20 1:38 PM

Visit the website for Broad Agency Announcement and

Untitled-1 1 6/15/20 11:44 AM

INNOVATORS THE STATE OF

6 The list Of course technology How the US lost its way

thought. Here’s many other things. Tales from the innovation

Krishna A crisis is no excuse not 64

and Arjuna to regulate tech Serve and replace

story Losing the winners’ game

The back page broader failures.

JA20_Contents.indd 4 6/8/20 1:36 PM

Editorial Corporate Consumer marketing MIT Technology Review

TR_masthead.indd 5 6/1/20 1:03 PM

JA20_Introduction.indd 6 6/5/20 4:38 PM

Where did all the money go?

1 Federal funding has been dropping

2.8 2000 2017

NATIONAL SCIENCE FOUNDATION NATIONAL SCIENCE FOUNDATION

JA20_Introduction.indd 8 6/5/20 1:01 PM

3 Spending on basic research has

600,000 ■ Development 1.6 8

NATIONAL SCIENCE FOUNDATION OECD US BUREAU OF LABOR STATISTICS

JA20_Introduction.indd 9 6/5/20 1:01 PM

JA20_Introduction.indd 10 6/5/20 1:01 PM

Apply today and get our latest research report free:

MA20 Global Panel Ad D1.indd 1 1/31/20 10:48 AM

as risks and prevent them from buying homes, getting

JA20_McIlwain-essay.indd 12 6/5/20 1:15 PM

jail, or secretly “disappear” them. Instead,

Reframing the problem

white racial power. murder of George Floyd by a white police

JA20_McIlwain-essay.indd 13 6/5/20 11:19 AM

MJ20 Podcasts D2.indd 1 3/31/20 11:01 AM

JA20_35-introduction.indd 15 6/8/20 1:02 PM

enzyme, Cas9. The tool, it turned out, was

JA20_35-inventors.indd 16 6/4/20 2:03 PM

Portrait by David Vintiner

JA20_35-inventors.indd 17 6/4/20 2:03 PM

“I think our goal right now is to have it

wave comes.” Age 28 | Country of birth: Ukraine

Her platform uses remote sensing

have it ready for the fall. For when the

JA20_35-inventors.indd 18 6/5/20 4:46 PM

Age 34 | Country of birth: Iran

She developed an AI-based system that

CHRISTINA BOVILLE olites in a patient can be

MIT, her team learned new

JA20_35-inventors.indd 19 6/4/20 2:04 PM

Her goal is to Age 32 | Country of birth: US