MIT Technology Review - March & April 2023

Everything design is USD $12.00 CAD $14.00 Mar/Apr 2023 Volume 126 Number 2 Designing drugs with AI Prosthetics that break the mold The future of ecological restoration

“ New IT” is an evolution built on smart devices, edge and cloud computing, 5G networks, and AI. And global biodiversity nonprofi t Island bridge 400 miles of Pacifi c Ocean. At Robinson Crusoe Island, one of the most remote places on Earth, it uses camera invasive species. Camera data used to be stored on a hard drive and periodically fl own to Santiago, Chile to process, taking as long as three months. Today, edge the island—time-savings that can save lives. “The Island Conservation team can process six months’ worth of visual data within just one week, enabling them to draw analytical insights within minutes instead of weeks,” Yuanqing says. If the pandemic taught technology leaders anything, it’s that public, private, hybrid, and multicloud computing is and development. “Normally, we wouldn’t think of tablets as life-saving equipment, but when emergency hospitals needed to be built during the Covid-19 outbreak, these played a critical role,” says Yuanqing. providing the fl exibility, scalability, and the co-existence of private, public, and The same Lenovo study found that cloud, software, and computing are key components for the future of a hybrid work environment, with 84% of respondents optimistic about the future of hybrid cloud. Connecting the essential components of a new IT architecture requires fast, effi cient, and customizable networking. The answer: 5G—the next generation of mobile wireless voice and data communication technology. The 2022 Lenovo study also found that 72% of CTOs see opportunities for their companies to use 5G multiaccess edge computing (MEC) even more with the demand for hybrid options dominating the workplace. “The popular hybrid work model that many companies have adopted over the last three years is only possible with a highspeed network,” says Yuanqing. By combining data, computing power, and sophisticated algorithms, AI can handle much more data much faster than a human worker, can be adjusted by users to accommodate change, can help users learn better processes, and can help anticipate risks such as cost overruns, accidents, and maintenance needs. Using multiple AI technologies and optimized algorithms, Lenovo Research created new that with some sixhour processes cut to 90 seconds. Lenovo estimates the AI solution improved order fulfi llment by 20% and productivity by 18%. Consider that a single PC order will launch a series of complex tasks across multiple production lines, and requires alignment of thousands of parameters, such as employee schedules, materials, production processes, and equipment statuses. Lenovo’s largest manufacturing base for PCs, LCFC Electronics, per year. While ing augmented reality headset with 3D video to give the user a realistic view of the work being done. The user’s head position controls the robot arm, and a handheld device controls movements. Edge computing helps data eliminate boundaries: Processing volumes of data can lead to performance issues. In response, many organizations are turning to edge computing, which processes data close to the source to enable fast and real-time analysis and response, while maintaining privacy and security requirements. “Edge computing allows data to be treated closer to where data is generated—directly at the edge site, lowering latency for faster response times, increased agility, and greater resilience,” says Yuanqing. For example, Kroger, one of the largest grocery chains in the United States, teamed with Lenovo and visual AI technology provider Everseen to build a system of secure self-checkout kiosks. AI servers capture unstructured data at each checkout from 20 high-resolution cameras. The system detects if an item is not scanned, and prompts the customer to rescan. It can also ping an associate’s mobile device. Since this requires enormous computing power, an edge solution processes the data near the source. “Over 75% of checkout errors can be corrected without employee intervention,” says Yuanqing. climate change with sustainable solutions. THE FIVE ELEMENTS OF NEW IT Although new technology and powerful applications are constantly emerging, Lenovo identifi es fi ve key components of a future-ready IT environment: smart devices, edge computing, cloud computing, high speed networks such as 5G, and AI. This defi nition resonates with technical leadership too, says Yuanqing, citing a 2022 Lenovo global research study of 500 chief technology offi cers in which four out of fi ve CTOs agree it “captures and describes the future of information communications technology (ICT) ‘extremely’ or ‘very well.’” Smart devices connect AI to human problems: According to Statista, the number of internet of things (IoT) devices worldwide will reach 29 billion IoT devices by 2030. IoT’s exponential growth—smart devices empowered by advanced sensors—provide a wide range of industries with competitive advantages. Manufacturers can use smart devices like robots to stand in for workers in dangerous or remote workspaces, and accelerate and automate assembly lines. For example, Lenovo’s Daystar Robot works remotely in real time using telepresence and teleoperation and learns tasks as it goes. The robot is operated by a streamBuilding the backbone for innovation, speed, and thriving humanity rom AI-powered platforms that can detect abnormal activities Fin supermarkets, to edge servers helping preserve biodiversity in remote locations, today’s technologies drive innovation in ways never before imaginable. “Innovation serves the purpose of making our life better, our work more productive, and our planet more sustainable,” says Yang Yuanqing, CEO and chairman of Lenovo. Technology leaders are reimagining an infrastructure where multiple technologies join to spur innovation in a secure, compliant, and user-friendly environment. Long gone are the days of “traditional IT and its client devices, servers, data centers, and on-premises applications,” says Yuanqing. He says traditional IT, shorthand for “information technology,” is being replaced by what Lenovo calls “new IT,” or “intelligent transformation.” Yuanquing explains that “The new IT enables digital transformation based on fi ve key elements: smart devices, edge computing, cloud computing, high-speed networks, and artifi cial intelligence. This new IT architecture can create countless opportunities.” This technology paradigm promises to support innovation and boost employee productivity, and also to power AI, revolutionize how enterprises use data, support business agility, and confront accounting for these large-scale calculations is a challenge for people, an AI engine can easily carry them out, and can fl exibly make real-time adjustments for broad or granular objectives. The AI solution’s autonomous learning ability also means the more it operates, the smarter it becomes. “This smart solution has also improved energy effi ciency and reduced greenhouse gas emissions by thousands of tons a year,” says Yuanqing. Technologies such as smart devices, edge computing, cloud computing, 5G, and AI are facilitating a shift from information technology to intelligent transformation. As always, while change surges ahead, technology executives must carefully consider the real-life outcomes of deploying new IT infrastructure. Security, compliance, and usability standards must still be upheld. “Environmental, social, and governance (ESG) goals must be a major consideration,” says Yuanqing. “In the future, every element of new IT architecture must incorporate ESG. When you assess the returns on innovation, it’s not just fi nancial payback but also social impact.” SPONSORED CONTENT

And global biodiversity nonprofi t Island Conservation uses edge computing to bridge 400 miles of Pacifi c Ocean. At Robinson Crusoe Island, one of the most remote places on Earth, it uses camera traps to document endangered and invasive species. Camera data used to be stored on a hard drive and periodically fl own to Santiago, Chile to process, taking as long as three months. Today, edge computing data centers process data on the island—time-savings that can save lives. “The Island Conservation team can process six months’ worth of visual data within just one week, enabling them to draw analytical insights within minutes instead of weeks,” Yuanqing says. Cloud computing provides connection: If the pandemic taught technology leaders anything, it’s that public, private, hybrid, and multicloud computing is imperative for fast and agile services and development. “Normally, we wouldn’t think of tablets as life-saving equipment, but when emergency hospitals needed to be built during the Covid-19 outbreak, these devices and innovative infrastructure played a critical role,” says Yuanqing. “In tough times, like the pandemic, it was new IT that kept us connected, productive, and engaged.“ He continues, “The public cloud became more popular by providing the fl exibility, scalability, and on-demand accessibility that we needed at the time. But, many enterprise applications and data are still running and stored in private cloud or on-prem data centers. In fact, we will continue to see the co-existence of private, public, and hybrid cloud for compute, storage, and network needs.” The same Lenovo study found that cloud, software, and computing are key components for the future of a hybrid work environment, with 84% of respondents optimistic about the future of hybrid cloud. 5G networks enable innovation and fl exibility: Connecting the essential components of a new IT architecture requires fast, effi cient, and customizable networking. The answer: 5G—the next generation of mobile wireless voice and data communication technology. The 2022 Lenovo study also found that 72% of CTOs see opportunities for their companies to use 5G multiaccess edge computing (MEC) even more with the demand for hybrid options dominating the workplace. “The popular hybrid work model that many companies have adopted over the last three years is only possible with a highspeed network,” says Yuanqing. AI tools mimic human intelligence to solve problems: By combining data, computing power, and sophisticated algorithms, AI can handle much more data much faster than a human worker, can be adjusted by users to accommodate change, can help users learn better processes, and can help anticipate risks such as cost overruns, accidents, and maintenance needs. Using multiple AI technologies and optimized algorithms, Lenovo Research created new processes for its manufacturing facility that dramatically improved production planning processes, with some sixhour processes cut to 90 seconds. Lenovo estimates the AI solution improved order fulfi llment by 20% and productivity by 18%. Consider that a single PC order will launch a series of complex tasks across multiple production lines, and requires alignment of thousands of parameters, such as employee schedules, materials, production processes, and equipment statuses. Lenovo’s largest manufacturing base for PCs, LCFC Electronics, processes up to 690,000 orders per year. While ing augmented reality headset with 3D video to give the user a realistic view of the work being done. The user’s head position controls the robot arm, and a handheld device controls movements. data can lead to performance issues. In response, many organizations are turning to edge computing, which processes data close to the source to enable fast and real-time analysis and response, while maintaining privacy and security requirements. “Edge computing allows data to be treated closer to where data is generated—directly at the edge site, lowering latency for faster response times, increased agility, and greater resilience,” says Yuanqing. For example, Kroger, one of the largest grocery chains in the United States, teamed with Lenovo and visual AI technology provider Everseen to build a system of secure self-checkout kiosks. AI servers capture unstructured data at each checkout from 20 high-resolution cameras. The system detects if an item is not scanned, and prompts the customer to rescan. It can also ping an associate’s mobile device. Since this requires enormous computing power, an edge solution processes the data near the source. “Over 75% of checkout errors can be corrected without employee intervention,” says Yuanqing. climate change with sustainable solutions. Although new technology and powerful applications are constantly emerging, Lenovo identifi es fi ve key components of a future-ready IT environment: smart devices, edge computing, cloud computing, high speed networks such as 5G, and AI. This defi nition resonates with technical leadership too, says Yuanqing, citing a 2022 Lenovo global research study of 500 chief technology offi cers in which four out of fi ve CTOs agree it “captures and describes the future of information communications technology (ICT) ‘extremely’ or ‘very well.’” According to Statista, the number of internet of things (IoT) devices worldwide will reach 29 billion IoT devices by 2030. IoT’s exponential growth—smart devices empowered by advanced sensors—provide a wide range of industries with competitive advantages. Manufacturers can use smart devices like robots to stand in for workers in dangerous or remote workspaces, and accelerate and automate assembly lines. For example, Lenovo’s Daystar Robot works and teleoperation and learns tasks as it goes. The robot is operated by a streamrom AI-powered platforms that in supermarkets, to edge servers helping preserve biodiversity in remote locations, today’s technologies drive innovation in ways never before imaginable. “Innovation serves the purpose of making our life better, our work more productive, and our planet more sustainable,” says . Technology leaders are reimagining an infrastructure where multiple technologies join to spur innovation in a secure, compliant, and user-friendly environment. Long gone are the days of “traditional IT and its client devices, servers, data centers, and on-premises applications,” says Yuanqing. He says traditional IT, shorthand for “information technology,” is being replaced by what Lenovo calls “new IT,” or “intelligent transformation.” Yuanquing explains that transformation based on fi ve key networks, and artifi cial intelligence. This technology paradigm promises to support innovation and boost employee productivity, and also to power AI, revolutionize how enterprises use data, support business agility, and confront accounting for these large-scale calculations is a challenge for people, an AI engine can easily carry them out, and can fl exibly make real-time adjustments for broad or granular objectives. The AI solution’s autonomous learning ability also means the more it operates, the smarter it becomes. “This smart solution has also improved energy effi ciency and reduced greenhouse gas emissions by thousands of tons a year,” says Yuanqing. A LOOK TO THE FUTURE Technologies such as smart devices, edge computing, cloud computing, 5G, and AI are facilitating a shift from information technology to intelligent transformation. “New IT is shaping the future in many innovative ways,” says Yuanqing. “In the future, the objects you work on, the colleagues you work with, the environment you work in, and the outcome you deliver might all be real or virtual, ranging from AI assistants and digital twins to the metaverse.” As always, while change surges ahead, technology executives must carefully consider the real-life outcomes of deploying new IT infrastructure. Security, compliance, and usability standards must still be upheld. “Environmental, social, and governance (ESG) goals must be a major consideration,” says Yuanqing. “In the future, every element of new IT architecture must incorporate ESG. When you assess the returns on innovation, it’s not just fi nancial payback but also social impact.” OF CTOs AGREE THESE 5 ELEMENTS CAPTURE & DESCRIBE THE FUTURE OF ICT ARCHITECTURE 87% SPONSORED CONTENT

G ood design has a habit of making things simple— sometimes too simple. You may look at the first iPod, for example, and marvel at its minimalist elegance without having to consider who designed it, where it was made and by whom, what materials it required, or even how long it would work. The ease of use and elegance of form erased the object from its context, an approach certainly not unique to Apple. As the American design educator Katherine McCoy observed in 1994, just seven years before the iPod’s release, “we have trained a profession that feels political or social concerns are either extraneous to our work or inappropriate,” despite the fact that “design is not a neutral value-free process. Design has operated this way in the world for a very long time. It still mostly does. While it is true, observes architect and designer Nicholas de Monchaux in his introduction to this issue, that design has accomplished much good in the world, “it has also shared responsibility for bringing us into our current ecological crisis; every new thing is perhaps not much better than the old thing.” Of course, we try to make new things that are better than what came before. But even big shifts are complicated. Take electric cars. They may not use fossil fuels but they come with their own trade-offs—a wide range of materials, from cobalt to copper to lithium, must be mined to build their batteries. Solving the resulting environmental challenges won’t begin to achieve another change that would likely do far more to reduce carbon emissions: figuring out how to get people to drive less. In her postmortem on design thinking on page 28, Rebecca Ackermann shows how, unintentionally, that iterative process for solving problems illustrated precisely the concerns voiced by McCoy. But Ackermann reports on a reckoning for design today and sees cause for optimism in new efforts to create design tools that are “capable of equitably serving diverse communities and solving diverse problems well into the future.” The design profession has—not for the first time and surely not for the last—been awakened to questions it hadn’t been asking before: Who is this for? Who is benefiting from it (and who or what might be harmed by it)? Who is being excluded? Have we explored the unintended consequences? Are we solving the right problem? These are just some of the questions we were thinking about when we were (yes) designing this issue, which features what you will see are not typical “design” stories. What they reveal is the astonishing breadth of what falls under the umbrella of design today. On page 36, Will Douglas Heaven delves into the use of AI automation for the design of new drugs, an approach that has the potential to deliver cheaper pharmaceuticals on a faster timeline. Matthew Ponsford explores the transformation happening on the outskirts of Mexico City, where the cancellation of a major international airport project created an opportunity to revive the nature and culture that once thrived there. Might this controversial wilderness point to the future of ecological design? John-Clark Levin’s fascinating commemoration of the 25th anniversary of the massively multiplayer online role-playing game Ultima Online, a precursor to the metaverse, shows how much the relative success or failure of design is contingent on human behavior. Do humans act the way the designer intended—or not? And on page 52, you’ll read about a movement in alternative prosthetics: creating devices that, instead of trying to mimic the appearance of a “normal” limb, make no attempt to blend in. Obstacles running the gamut from conformist thinking to cost have inspired designers to forge a new path, one that may, writes Joanna Thompson, “help prosthetics users wrest back control of their own image and feel more empowered, while simultaneously breaking down some of the stigma around disability and limb difference.” If we accept that everything is design, and by extension that everyone is a designer, then our expectations for the discipline may have been unrealistic, even misguided. “It is no exaggeration to say that designers are engaged in nothing less than the manufacture of contemporary reality,” wrote designer Rick Poynor in 1999. What might be different now is we recognize the responsibility that comes with being a part of that process. 02 From the editor By Allison Arieff If design is everything, is it anything? Allison Arieff is editorial director of print for MIT Technology Review.

www.technologyreview.com/thecloudhub We made 3 organizations, 3 cultures, 3 ERP systems and 3000 applications work in unison.

04 2 From the editor THE DOWNLOAD 11 The consequences of e-commerce, geoengineering bans, Cell Painting, better tsunami detection, and reasons for climate optimism. Plus, this month’s job title of the future: Director of Species Restoration EXPLAINED 20 Explained: How do lithium-ion EV batteries work? By Patrick Sisson PROFILE 22 Tackling diversity in Germany’s tech capital Nakeema Stefflbauer is bringing women from underrepresented groups into the Berlin tech scene. By Gouri Sharma 72 Rust never sleeps How a ragtag international group of coders brought a much-loved and fastgrowing new programming language into the world. By Clive Thompson 78 Power pop K-pop fans are getting political, using the campaign techniques they’ve perfected promoting their favorite bands online. By Soo Youn 82 Welcome to the strange, surprising world of Ultima Online What a decades-old virtual kingdom can teach us about the future of the metaverse. By John-Clark Levin ARCHIVE 88 Graphic designer Jacqueline Casey helped define MIT’s distinctive visual style. Front The design issue Back 25 What is design? When we unpdck its current medning, we mdy find thdt we wdnt—dnd need—to retool the word yet dgdin. BY NICHOLAS DE MONCHAUX 28 A postmortem on design thinking Empdthy fdiled to fix the world. Whdt now? BY REBECCA ACKERMANN 36 Drugs by design AI dutomdtion is being deployed throughout the drug development pipeline, opening up the possibility of chedper phdrmdceuticdls on d fdster timeline. BY WILL DOUGLAS HEAVEN 42 Electromagnetic ruins Ohio’s dncient mounds dnd the future of drchdeology. BY GEOFF MANAUGH 52 Prosthetics that break the mold Prosthetics designers dre pushing the limits of functiondlity dnd desthetics to help people feel more comfortdble in their own skin. BY JOANNA THOMPSON 60 The return of Lake Texcoco Mexico cdnceled d $13 billion dirport. It’s now spending nedrly $1 billion to turn it bdck into d swdmp. Does this controversidl wilderness point to the future of ecologicdl design? BY MATTHEW PONSFORD “I want to make something that not only changes things, but changes things without screwing everything else up.”–p. 28 Cover illustration by Israel Vargas COVER ILLUSTRATION SOURCES: WIKIMEDIA (BAUHAUS, IPOD); ALESSI (JUICER); AMAZON (ECHO DOT); GETTY (TEXTILE, ESPRESSO); NEST (THERMOSTAT); MAEDAAKIHIKO/WIKIMEDIA (TRAIN); IROBOT (ROOMBA); TESLA (MODEL X); THE MUSEUM AT FIT (CHANEL SUIT); ORAL-B (TOOTHBRUSH); IDUKE/WIKIMEDIA (SUBURBS); JOHN ATHERTON WIKIMEDIA (TV)

R&D FUNDING PROGRAM The National Reconnaissance Offi ce Director’s Innovation Initiative (DII) funds cutting-edge scientifi c research in a highrisk, 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. www.nro.gov/innovate [email protected] TRANSFORM OUR FUTURE TRANSFORM OUR FUTURE INNOVATE AWARDS UP TO $500K APPLICATIONS OPEN 20 MARCH – 5 MAY AREAS OF INTEREST Apertures Communications Remote Sensing Satellite System Design Sense-Making Other Disruptive Concepts & Technology

06 Masthead Editorial Editsr in chief Mat Honan Executive editsr, speratisns Amy Nordrum Executive editsr, newsrssm Niall Firth Editsrial directsr, print Allison Arieff Editsrial directsr, audis and live jsurnalism Jennifer Strong Editsr at large David Rotman Science editsr Mary Beth Griggs News editsr Charlotte Jee Features and investigatisns editsr Amanda Silverman Managing editsr Timothy Maher Csmmissisning editsr Rachel Courtland Senisr editsr, MIT News Alice Dragoon Senisr editsr, bismedicine Antonio Regalado Senisr editsr, climate and energy James Temple Senisr editsr, digital culture Abby Ohlheiser Senisr editsr, AI Will Douglas Heaven Psdcast prsducer Anthony Green Senisr repsrters Tanya Basu (humans and technology) Eileen Guo (features and investigations) Jessica Hamzelou (biomedicine) Melissa Heikkilä (AI) Tate Ryan-Mosley (tech policy) Repsrters Casey Crownhart (climate and energy) Rhiannon Williams (news) Zeyi Yang (China and East Asia) Cspy chief Linda Lowenthal Senisr audience engagement editsr Abby Ivory-Ganja Audience engagement editsr Juliet Beauchamp Creative directsr, print Eric Mongeon Digital visuals editsr Stephanie Arnett Corporate Chief executive sfficer and publisher Elizabeth Bramson-Boudreau Finance and Operations Chief financial sfficer, head sf speratisns Enejda Xheblati General ledger manager Olivia Male Accsuntant Anduela Tabaku Human ressurces directsr Alyssa Rousseau Manager sf infsrmatisn technslsgy Colby Wheeler Data analytics manager Christopher Doumas Office manager Linda Cardinal Technology Chief technslsgy sfficer Drake Martinet Vice president, prsduct Mariya Sitnova Senisr ssftware engineer Molly Frey Asssciate prsduct manager Allison Chase Digital brand designer Vichhika Tep Events Senisr vice president, events and strategic partnerships Amy Lammers Directsr sf event csntent and experiences Brian Bryson Head sf internatisnal and custsm events Marcy Rizzo Senisr event csntent prsducer Erin Underwood Directsr sf events Nicole Silva Event speratisns manager Elana Wilner Manager sf strategic partnerships Madeleine Frasca Williams Event cssrdinatsr Bo Richardson Consumer marketing Vice president, marketing and csnsumer revenue Alison Papalia Directsr sf retentisn marketing Taylor Puskaric Directsr sf acquisitisn marketing Alliya Samhat Directsr sf event marketing Nina Mehta Email marketing manager Tuong-Chau Cai Circulatisn and print prsductisn manager Tim Borton Advertising sales Senisr vice president, sales and brand partnerships Andrew Hendler [email protected] 201-993-8794 Asssciate vice president, integrated marketing and brand Caitlin Bergmann [email protected] Executive directsr, brand partnerships Marii Sebahar [email protected] 415-416-9140 Executive directsr, brand partnerships Kristin Ingram [email protected] 415-509-1910 Executive directsr, brand partnerships Stephanie Clement stephanie.clement@ technologyreview.com 214-339-6115 Executive directsr, sales and brand partnerships Debbie Hanley [email protected] 214-282-2727 Senisr directsr, brand partnerships Ian Keller [email protected] 203-858-3396 Senisr directsr, brand partnerships Miles Weiner [email protected] 617-475-8078 Senior director, digital strategy, planning, and ad ops Katie Payne [email protected] Digital sales strategy manager Casey Sullivan [email protected] Media kit www.technologyreview.com/media MIT Technology Review Insights and international Vice president, Insights and internatisnal Nicola Crepaldi Glsbal directsr sf custsm csntent Laurel Ruma Senisr manager sf licensing Ted Hu Senisr editsr, custsm csntent Michelle Brosnahan Senisr editsr, custsm csntent Kwee Chuan Yeo Editsr, custsm csntent Teresa Elsey Senisr prsject manager Martha Leibs Prsject manager Natasha Conteh Directsr sf partnerships, Eurspe Emily Kutchinsky Directsr sf partnerships, Asia Marcus Ulvne Board of directors Cynthia Barnhart, Cochair Alan Spoon, Cochair Lara Boro Peter J. Caruso II, Esq. Whitney Espich Sanjay E. Sarma David Schmittlein Glen Shor Customer service and subscription inquiries Natisnal 877-479-6505 Internatisnal 847-559-7313 Email [email protected] Web www.technologyreview.com/ customerservice Reprints [email protected] 877-652-5295 Licensing and permissisns [email protected] MIT Technology Review 196 Broadway, 3rd Floor Cambridge, MA 02139 617-475-8000 Our in-depth reporting reveals what’s going on now to prepare you for what’s coming next. 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.

Great Satellites launch more than once Investment into research and development in the UK is forging new paths for the global space industry. -isruptive new technologies are being developed by UK firms, such as the world’s first returnable and re-launching satellite platform. To see things differently, choose the UK. great.gov.uk/r-and-d Space Forge The ForgeStar™ Platform

Trusted The essential emerging technology event for leaders who need to understand the impact of tomorrow’s technology, today. April 30 – May 1, 2023 Online or in person @ MIT FutureComputeMIT.com Compete with compute. Future Compute & EmTech Digital run

ed insights. Only from MIT Technology Review The leading executive briefing on artificial intelligence, its implementation, and impact on business and society. May 2 – 3, 2023 Online or in person @ MIT EmTechDigital.com Capitalize on artificial intelligence. back to back. Attend both and save!

SEIZE THE OPPORTUNITIES EXPOSED BY CUTTING-EDGE TECHNOLOGY. Digital transformation is a journey, not a destination. S U B S C R I B E R S S A V E 1 0 % W I T H C O D E P R I N T M A 2 3 A T EmTechNext.com June 13 – 14, 2023 Online or in person @ MIT

CHRIS PIASCIK 11 multitrillion-dollar market led by China, and one of the few tech industries in the world in which China has been spearheading innovations. For years, people have been asking whether the US will ever catch up. Then Shein entered the public’s attention and changed things again. Founded in China in 2008, Shein has become incredibly popular far beyond the country’s borders, with a valuation of about $100 billion. Young people on almost all continents are inspired by TikTok and YouTube influencers doing “Shein hauls”—buying dozens of garments and other items in one go and dumping everything onto the bed before trying them on for the camera, one by one. The sheer contrast between the tall pile of clothes and the low total price tag makes for a great social media spectacle. For me, the selling point came when a friend here, also from China, told me: “It feels just like Taobao.” I went on my first Shein journey in August 2021. And guess what I bought? Yeah, phone cases. I bought six phone cases and one wireless-earbud case for a total of $12.50—taxes and shipping included. They were fun and brightly colored, and the quality was not bad. I told myself I had finally found a way to replicate my cheap shopping experience here, thanks to a Chinese company that wants to expand to the world. I suspect for many recent Chinese immigrants like me, shopping on Shein is like microdosing Taobao or the Chinese e-commerce system more broadly. Whenever I’m on Shein, suddenly I feel surrounded by the familiar—both because of the price points and because there are dozens of pages of products Why my bittersweet relationship with Shein had to end Reflecting on my desire for Chinese-style e-commerce platforms. By Zeyi Yang To be honest, I’ve been missing the online shopping experience in China since I moved to the US four years ago. I grew up in China at the same time that Taobao, a popular ecommerce platform, inserted itself into the center of everyday life. Whatever common, luxury, niche, or handmade products you wanted, you could always find them online on Taobao, and at a cheaper price than in brick-and-mortar stores. It really is the place of abundance. So when I noticed Shein becoming mainstream in the US over the past few years, I thought, Great! I finally have a Taobao replacement! But somewhere along the way, I started questioning why I enjoy this particular kind of shopping, and also what it means for an e-commerce platform to offer endless deals. Before I moved in 2018, I made a purchase on Taobao: five phone cases. I bought them precisely because I’d heard that small things like phone cases were much more expensive on the other side of the Pacific. I was not wrong. I paid roughly $15 total, and that included shipping from four different sellers. In the US I could easily spend $50, if not more, for the same. While it was difficult for me to say goodbye to Taobao, I had to accept the new reality of online shopping in the US: things are more expensive, there are not as many options, and it takes longer to ship everything. Taobao is just one of the handful of creations from Chinese e-commerce giants like Alibaba, JD, and Pinduoduo that learned to combine the country’s traditional advantage of low manufacturing and transportation costs with the fast pace of tech innovations in mobile payment, recommendation algorithms, and targeted marketing. As a result, e-commerce has become a The Download

COURTESY OF THE PUBLISHERS Book reviews 12 The Download that I can never get to the bottom of. But after a while, as the not-so-contradictory feelings of novelty and nostalgia wore off, I stopped looking for deals on Shein. I started to ask myself: Did I really need that many phone cases?! To be fair, there are Shein purchases that I’ve really enjoyed, like a $2 nylon watchband that feels better than my original Apple Watch band. I also think people should be able to choose quantity and price over quality, because the idea of demanding that people only buy premium products feels unrealistic and patronizing. But as it turns out, I’ve finally started to see through the illusion of Shein-like platforms. To get these occasional incredible deals, you are encouraged to shop much more than is necessary or even reasonable. This illusion has worked for a long time and for a lot of people—including me!—but it’s become harder and harder to ignore the environmental consequences of my purchases, and the ways in which platforms trick people into buying more and more. And I don’t think I’m the only one experiencing that awakening. Broadly speaking, I think society is slowly but surely shifting toward recognizing the climate impact of mass-produced cheap goods (and, maybe even more slowly, the related costs of cheap human labor). While these conversations have yet to happen as widely and furiously in China, companies like Taobao and Shein will inevitably have to answer the question of whether their business model is sustainable for everyone—or only for themselves. The Chinese e-commerce industry, which had grown at spectacular speed for a decade, now needs to contend with bubbling public pressure regarding the environment and the economic pressure of a recession on the horizon. So where are they heading from here? There’s certainly a lot of soul-searching for the industry to do. And I’m doing some soul-searching of my own. ■ Visit www.technologyreview.com to read the full story. We Are Electric: Inside the 200-Year Hunt for Our Body’s Bioelectric Code, and What the Future Holds By Sally Adee (Hachette, 2023) Electrical signals are “pressed into service by every cell in your body,” explains science writer Sally Adee. In exploring the science of bioelectricity, a field that is still very much emerging from a long period of skepticism, she shows how cracking the body’s electric code could open up new realms of medicine. Blood Money: The Story of Life, Death, and Profit Inside America’s Blood Industry By Kathleen McLaughlin (Simon & Schuster, 2023) For years, American journalist Kathleen McLaughlin smuggled blood plasma into China, where she lived and reported, to manage a rare chronic condition. Building on her own experiences, Blood Money is an engrossing look into the globe-spanning business of selling blood in the United States and China—and the inequality that creates such a market. Arcade Game Typography by Toshi Omagari (Read-Only Memory, 2019) “I was oddly grateful of the lack of a wife or girlfriend during this period,” says Omagari, describing the process of writing this love letter to Galaxian, Monster Bash, Quasar, and all the other power- and colorpacked typefaces of arcade games from the 1970s, ’80s, and ’90s. His obsession shows on every page. You Are Not Expected to Understand This”: How 26 Lines of Code Changed the World Edited by Torie Bosch (Princeton University Press, 2022) Is code impenetrable to so many because it has to be or because very human decisions made it so? That’s the question guiding these 26 short, sharp essays, the topics of which range from French tapestry to the first-ever line of code, an apology for writing the JavaScript code that “made the world a measurably worse place” (because it led to the pop-up ad), and a look at how an algorithm developed in the 1960s led to modern racial profiling. ■ “

SHAWN HAZEN The Download 13 An American entrepreneur’s crude solar geoengineering effort in Baja California, first reported by MIT Technology Review in late December, provoked widespread criticism. Within three weeks, the Mexican government announced plans to ban related experiments. Luke Iseman, previously a director of hardware at Y Combinator and the cofounder of a geoengineering startup, says he added a few grams of sulfur dioxide into a pair of weather balloons and launched them from an unspecified site somewhere on the Mexican peninsula last spring. He says he intended for the balloons to reach the stratosphere and burst under pressure there, releasing the particles into the open air. Scientists believe that spraying sulfur dioxide or other reflective particles into the stratosphere in sufficient quantities might be able to offset some level of global warming, mimicking the cooling effect from major volcanic eruptions in the past. But it’s a controversial field, given the unknowns about potential side effects, fears that even discussing the possibility could undermine the urgency to address the root causes of climate change, and other thorny questions. Iseman acknowledged to MIT Technology Review, and other outlets reporting on the effort, that he didn’t seek scientific or government approval before moving forward with the balloon launches. He subsequently cofounded the startup, Make Sunsets, to commercialize the concept, offering to sell “cooling credits” for particles released during future balloon launches. Iseman previously said he started the company to combat the rising dangers of climate change. He had hoped forging ahead with the launches would help push forward a scientific field that, amid public criticism, has repeatedly encountered serious challenges to carrying out smallscale field experiments. But on January 13, Mexico’s Ministry of Environment and Natural Resources announced that the government will prohibit and, where appropriate, halt any solar geoengineering experiments within the country. The agency noted that Make Sunsets’s launches were done without notice or consent. It said the prohibition was motivated by the risks of geoengineering, the lack of international agreements supervising such efforts, and the need to protect communities and the environment. Mexico may be one of the first nations, if not the first, to announce such an explicit ban on experiments, although many nations have existing environmental regulations and other policies that could restrict certain practices. Iseman, who didn’t respond to an inquiry from MIT Technology Review, told The Verge that future launches are “indefinitely on hold.” He said to the Wall Street Journal that he was “surprised by the speed and scope of the response” and had “expected and hoped for dialogue.” But others weren’t surprised. Shuchi Talati, a scholar in residence at American University who is forming a nonprofit focused on governance and justice in solar geoengineering, warned in MIT Technology Review’s original piece that Make Sunsets’s actions could have a chilling effect on the field. She said the unauthorized effort could diminish government support for geoengineering research and amplify demands to restrict experiments. The nation’s response “highlights the reckless way in which this company acted,” Talati said in an email. “To go to another country and conduct something akin to experimentation without consultation or engagement is unacceptable.” Geoengineering researchers stress that they want to explore the potential of the technology because it could save lives— possibly many, many lives as heat waves, famines, wildfires, and other extreme weather events grow more common and severe in the coming decades. Some fear that Make Sunsets’s rudimentary balloon launches and attempts at commercialization could distort the perception of the field among the public and policymakers, and dampen support for research. “I think there’s a danger of painting serious scientists and careful experiments with the same brush as some dude that released some weather balloon and tried to make a buck off of it,” says Peter Irvine, a lecturer in climate change and solar geoengineering at University College London. “There are many of us who think seriously studying this idea … is worth doing, because it looks like it has the potential to substantially decrease the risk of climate change,” he adds. “We shouldn’t throw the baby out with the bathwater, and that’s the worry.” ■ A startup’s bid to commercialize geoengineering provoked a clampdown in Mexico The nation announced plans to ban experiments in the field after an entrepreneur said he’d launched weather balloons that might have released reflective particles into the atmosphere. By James Temple

COURTESY OF JUMP-CELL PAINTING CONSORTIUM 14 The Download Cell Painting speeds drug discovery Appublic-privatepconsortiumpp haspreleasedpaphugepcollectionpp ofpimage-basedpcellpprofiles. By Esther Landhuis years to come. The method, known as Cell Painting, impressed scientists at several pharmaceutical companies—so much that they launched a consortium and pooled resources, using the approach to create a massive data set that they began releasing to the public in November. The JUMP–Cell Painting Consortium, as it’s called, hopes the database will accelerate drug discovery by helping researchers identify promising compounds and get a better sense of what they do and what sorts of side effects they might have before the molecules get tested in animals or people. Cell Painting uses up to six fluorescent dyes to light up major components of the cell, such as the nucleus and mitochondria. A microscope snaps images of the various stains, and software measures morphological features like size, shape, intensity, and texture, creating an image-based profile of the sample. It is “just about the simplest imaging assay you can manage,” says computational biologist Anne Carpenter, who developed the method and co-leads the Broad Institute lab with Shantanu Singh. “Our mission was to choose the absolute cheapest, easiest dyes.” Beyond ease of use, the power of Cell Painting lies in the sheer volume of data that comes from one experiment. The newly released database contains images of cells responding to more than 140,000 perturbations—either a drug treatment or some other modification that turns a gene’s activity up or down. Using this data set, Carpenter and some of her colleagues found a dozen compounds that seem to affect the same structures that are influenced by a key gene involved in a fast-growing muscle cancer. Rather than putting hundreds of samples through multiple rounds of wet-lab experiments, the Broad researchers came up with the drug list several years ago by typing the name of the gene into the database. “It’s a totally different approach that has a lot fewer steps and is a lot less costly,” says T.S. Karin Eisinger, a biologist at the University of Pennsylvania who studies that particular muscle cancer. Her team worked with Carpenter’s to validate the compounds in wet-lab tests, and the two scientists are launching a company to further develop the most promising candidates. Others are a bit further along: Recursion Pharmaceuticals, a company in Salt Lake City for which Carpenter is an advisor, has already launched five clinical trials to test drug candidates identified using a version of Cell Painting. As it wraps up its public release, consortium members are gearing up to work with the Health and Environmental Sciences Institute, based in Washington, DC, to see if they can pair results from Cell Painting with other data to predict the toxicity of pharmaceuticals and agrochemicals. ■ Esther Landhuis is a science and health journalist based in the San Francisco Bay Area. One of the earliest stages in the process of identifying a potential new drug is to expose cells to the compound in a lab dish and scour microscope images to see the effects. Biologists who do this work tend to focus on a few select features that could indicate the drug is working—a cluster of fluorescently labeled proteins, for example, or a decrease in the number of dividing cells. The strategy is tedious and time-consuming, and it often fails because researchers aren’t sure what to look for or where in the cell to look. Now some are embracing a new paradigm: Measure everything, ask questions later. This motto drives a lab at Harvard and MIT’s Broad Institute, where researchers have developed a method for generating a treasure trove of information on a cell’s inner workings that they can sift for

TELEGEOGRAPHY The Download 15 Building better tsunami detection Telecompcompaniesphaveplongpresistedplettingp scientificpsensorsppiggybackponptheirpsubseap cables—untilpnow. By Christian Elliott The residents of Vanuatu, a clutch of islands in the South Pacific, are no strangers to flooding. The ocean floor around them is frequently shaken by tsunami-triggering earthquakes. Some advance warning could give residents enough time to get to higher ground before tsunamis strike, saving lives. But the world’s 65 active deep-ocean buoys, which are designed to detect the waves, are too sparsely distributed to provide that type of warning for Vanuatu. The Joint Task Force for Science Monitoring and Reliable Telecommunications (SMART) Subsea Cables, a United Nations initiative, aims to solve that problem by equipping new commercial undersea telecom cables with simple sensors that measure pressure, acceleration, and temperature. The sensors could be added to the fiber-optic cables’ signal repeaters—the watertight cylinders full of equipment that are used to amplify signals every 50 kilometers or so. With cables providing for the sensors’ power and data transfer needs, scientists could collect information about the seafloor at an unprecedented scale—and pass on data about potential tsunamis far faster than is currently possible. Adding sensors to undersea cables isn’t a new idea. Bruce Howe at the University of Hawaii, for example, who chairs the task force, operates the deepest science observatory in the world using an abandoned telecom cable located 100 kilometers north of Oahu. But convincing the $5 billion-a-year subsea telecommunications industry to integrate scientific sensors into the expensive hardware it installs has been an uphill battle for a decade, Howe says. A big part of the challenge is that a repeater needs to be pressurized against conditions kilometers underwater. Adding external sensors that must be powered by—and communicate with—the repeater complicates the design. But last year Subsea Data Systems, a startup with funding from the US National Science Foundation, built a prototype repeater that showed it could be done. This year the technology is scheduled to have its first true “wet” demonstration when three test repeaters are deployed off the coast of Sicily. Governments—and companies—are starting to get on board. The major telecom cable company Alcatel recently announced it would have SMART cable technology ready by 2025. That same year, Portugal plans to begin work on CAM, a €150 million SMART cable project to connect Lisbon with the islands of Madeira and the Azores. The European Union has designated €100 million for digital connectivity infrastructure, including these types of cable projects. These are heartening developments for scientists interested in expanding our ability to study the changing ocean, something that is currently done mostly from space and by research ships. And if the technology comes to Vanuatu and New Caledonia, a neighboring island nation, it could mean a big change in public safety. The two small countries are separated by an area where one section of ocean floor is actively diving beneath another, causing those frequent earthquakes and tsunamis. Residents may have a few minutes, or even just seconds, to respond to a tsunami alert. According to new modeling by the task force, presented at the American Geophysical Union conference in Chicago in December, a SMART cable across this “subduction zone” could extend that lead time to 12 minutes. It would also provide a second high-speed connection to the outside world for Vanuatu, reducing the risk of communication blackouts like the one that occurred last year in Tonga when a volcanic eruption severed that country’s only telecom cable. “If we can give a community even five or 10 minutes of additional time, that can make a huge difference,” says Laura Kong, a member of the task force and director of the International Tsunami Information Center, a joint effort of UNESCO and the US National Oceanic and Atmospheric Administration. Researchers have high hopes—and big plans—for SMART cables. In addition to the idea of a Vanuatu–New Caledonia cable, they are proposing projects in New Zealand, the Mediterranean, Scandinavia, and even Antarctica. “This is a first step in achieving a long-term vision of instrumenting the ocean seafloor for climate and early-warning purposes,” says Howe. “This is the first time the deep ocean would be sort of opened up in this way.” ■ Christian Elliott is a freelance science journalist based in Chicago. The Submarine Cable Map

16 The Download A few pieces of good news on climate change (and a reality check) Theppathpaheadpforpclimatepactionpispnarrow,p butpapcloseplookpatpemissionspdatapprovidesp apfewpreasonsptopbepoptimistic. By Casey Crownhart Emissions from fossil-fuel sources were higher than ever in 2022, according to data from the Global Carbon Project. Global growth year-over-year was just over 1%, continuing a rebound from a 2020 low caused by the covid-19 pandemic. Overall, emissions have doubled in about the last 40 years. But while emissions grew globally, in many countries they have already plateaued or begun to decrease. US emissions peaked in 2005 and have declined by just over 10% since then. Russia, Japan, and the European Union have also seen emissions level off. Global emissions are expected to reach their peak in about 2025, according to the International Energy Agency. Reaching maximum annual emissions is a significant milestone, the first step in turning the metaphorical ship around for greenhouse gases. But emissions are still growing in some countries, including China (the world’s current leading emitter) and India, both of which have growing populations and economies. China’s increase has been especially sharp, with emissions roughly doubling over the past 15 years. China’s government has pledged that the country will reach its emissions peak by 2030 and achieve net-zero emissions before 2060. The peak could come even sooner, in 2025 or before, according to an analysis by CarbonBrief. The nation is deploying renewables at record speed, roughly quadrupling installations over the past decade. India’s emissions increase is more moderate than China’s, but the country will likely see growth continue until 2040 or 2050. For now, though, its total emissions are far less than those of China and the US, and it is behind most other countries in per capita emissions. When it comes to climate, the picture can look bleak. Emissions of the greenhouse gases that cause climate change reached a new peak in 2022, according to early estimates. And climate disasters seem to be hitting at a breakneck pace. In 2022, the world experienced record heat waves in China and Europe, and devastating floods in Pakistan killed over 1,000 people and displaced millions. But a close look at energy and emissions data around the world shows that there are a few bright spots of good news, and a lot of potential progress ahead. For example, renewable sources make up a growing fraction of the energy supply, and they’re getting cheaper every year. Countries are setting new targets for emissions reductions, and unprecedented public investments could unlock more technological advances. There are at least a few reasons to be hopeful: While emissions reached new heights in 2022, the peak is in sight. Annual emissions from fossil fuels and industry (gigatons) Global emissions from fossil fuels, flaring, and cement productions (gigatons) 1950 1960 1970 1980 1990 2000 2010 2020 SOURCE: GLOBAL CARBON PROJECT, CASEY CROWNHART/MIT TECHNOLOGY REVIEW 0 40 30 20 10 5 35 25 15 US EU China India Russia Japan 10 9 8 7 6 5 4 3 2 1 0 12 11 1950 1960 1970 1980 1990 2000 2010 2020

The Download 17 Economic growth is becoming less dependent on fossil fuels Emissions have tended to increase with economic growth, but in the future, progress on emissions won’t necessarily require sacrificing economic gains. As renewable energy is more widely deployed and technical improvements drive efficiency, economic growth may be possible without a proportional rise in climate pollution. Getting to the bottom of which countries have contributed most to climate change is complicated, but a few pieces of data can help. Some nations have already begun to cut emissions while maintaining economic growth. Helping developing nations to do the same will be vital. Globally, the carbon intensity of economic growth is dropping over time, meaning the carbon emissions associated with a given level of economic activity have decreased. This is true globally, as well as for large economies like the US and EU. The trend is most obvious in China, where the carbon intensity of the economy has dropped by about 40% since 2000. But China’s carbon intensity is still higher than that of most other large nations. And progress has slowed, largely because of the high proportion of coal in the country’s energy mix today— about 60%, as of 2021. The reality check: climate progress needs to happen even faster While emissions are leveling off or dropping in some parts of the world, even the countries that are making progress largely aren’t doing so fast enough to reach international climate goals. The Paris Agreement, an international climate treaty adopted in 2015, set a target to keep warming at less than 2 °C over preindustrial levels, or ideally less than 1.5 °C. From climate models, researchers have estimated the limits to total greenhouse-gas emissions needed to hit these targets. The concept is called the global carbon budget, and we’ve nearly spent it all. If we had started emission cuts sooner, our carbon budget might have stretched farther into the future, allowing for more gradual cuts. But now, in order to keep warming under 1.5 °C globally given historical emissions, we need to reach net zero by 2050; by 2030 the world’s emissions will need to be cut roughly in half. And even that might not be enough. Keeping warming under 1.5 °C is possible, though the goal is slipping out of reach, given that global surface temperatures have already increased by about 1.1 °C since before 1900. How much more temperatures rise in the future will be a function of emissions, and the sooner significant cuts happen, the more likely we are to keep warming close to the 1.5 °C target. It’s clear that building renewable energy and finding other ways to cut emissions can slow climate change. Whether you see it as good news or bad news, the future will be dictated by the world’s actions today and in the near future. ■ Subscribe to our climate newsletter, The Spark, at www.technologyreview.com Carbon intensity of GDP (metric tons of carbon dioxide per $1,000) Projected emission levels needed to keep warming under 1.5 °C (gigatons) LEFT: INTERNATIONAL ENERGY AGENCY, CROWNHART/MIT TECHNOLOGY REVIEW; RIGHT: GLOBAL CARBON PROJECT, CROWNHART/MIT TECHNOLOGY REVIEW US EU China India Japan World 1980 1985 1990 1995 2000 2005 2015 2010 2020 0 2 1 0.8 0.6 0.4 0.2 1.8 1.6 1.4 1.2 1980 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 0 40 30 20 10 5 45 35 25 15 Historical Emissions Start in 2000 Start in 2020 Start in 2010

JAN BUCHCZIK 18 The Download A company called Colossal Biosciences aims to return lost species to the planet. Sara Ord, 28, whose background is in chemical and biomolecular engineering, is tasked with figuring out how to bring back the carnivorous marsupial known as the thylacine, or Tasmanian tiger, the last of which died in 1936. Qualifications needed: A love of animals, a background in science, and plenty of hours watching Discovery Channel. “I grew up with dogs, cats—we had a tortoise, hamsters, fish,” says Ord. “De-extinction is kind of my dream job.” Current challenge: The sharply-clawed thylacine is gone, but a distant, mouse-size cousin, called the dunnart, remains. Ord leads a team of 12 that is trying to determine how to genetically modify dunnarts so that they start to resemble thylacines. Scientists do know what thylacine DNA should look like. That’s because of the so called “miracle pup”—a youngster someone plopped in a jar of alcohol more than 100 years ago. That specimen yielded an accurate gene map. Job perk: Getting to focus on a creature that used to be at the top of a food chain—and explain to the public why such “apex predators” are so important for healthy ecosystems. “This is a species that we drove to extinction,” says Ord. “Bringing it back will hold tremendous value, as well as righting a wrong.” Commercial prospects: Colossal hopes to generate a profit by spinning off technology, like genomics software, as well as cashing in on media attention. “One of our products is the story, right?” says Ord. If Colossal is successful in re-creating a thylacine by 2025, as it plans, the company might sell tickets to see it. How much would Ord pay to catch a glimpse of one? “Well, I am spending hours and hours of my life on this,” she says, “so it would be worth all the money in the world.” ■ Job titles of the future: Director of Species Restoration By Antonio Regalado

20 Explained T he batteries propelling electric vehicles have quickly become the most crucial component, and expense, for a new generation of cars and trucks. They represent not only the potential for cleaner transportation but also broad shifts in geopolitical power, industrial dominance, and environmental protection. According to recent predictions, EVs will make up just over half of new passenger car sales in the US by 2030. One estimate suggests that the potential growth of the global battery market could require 90 more facilities the size of the Tesla Gigafactory to be built over the next decade. Lithium-ion batteries, also found in smartphones, power the vast majority of electric vehicles. Lithium is very reactive, and batteries made with it can hold high voltage and exceptional charge, making for an efficient, dense form of energy storage. These batteries are expected to remain dominant in EVs for the foreseeable future thanks to plunging costs and improvements in performance. Right now, electric-car batteries typically weigh around 1,000 pounds, cost around $15,000 to manufacture, and have enough power to run a typical home for a few days. While their charging capacity degrades over time, they should last 10 to 20 years. Each battery is a densely packed collection of hundreds, even thousands, of slightly mushy lithium-ion electrochemical cells, usually shaped like cylinders or pouches. Each cell consists of a positive cathode (which typically contains metal oxides made from nickel, manganese, and cobalt); a negative, graphitebased anode; and a liquid solution in the middle, called an electrolyte. This is where lithium’s reactivity comes into play; its loosely held outer electron can easily be split off, leaving a lithium ion (the atom sans its outer electron). The cell basically works by ping-ponging these ions and electrons back and forth. During the charging cycle, an electric current introduced via an external source separates the electrons from the lithium atoms in the cathode. The electrons flow around an outside circuit to the anode—which is typically composed of graphite, a cheap, energy-dense, and long-lasting material that excels at storing energy—while the ionized lithium atoms flow to the anode through the electrolyte and are reunited with their electrons. Electric vehicles need them now, but the search is on for cheaper, more sustainable alternatives. By Patrick Sisson Illustration by Lorenzo Petrantoni How lithiumion EV batteries work

The Download Explained 21 During discharge cycles, the process reverses. Lithium atoms in the anode get separated from their electrons again; the ions pass through the electrolyte; and the electrons flow through the outside circuit, which powers the motor. EV expansion has created voracious demand for the minerals required to make batteries. The price of lithium carbonate, the compound from which lithium is extracted, stayed relatively steady between 2010 and 2020 but shot up nearly tenfold between 2020 and 2022, spurring new investments across the globe. More than a dozen battery plants and numerous potential mining projects are in development in the US alone. But the quest for raw materials comes with extensive environmental, political, and social costs. The vast majority of cobalt, a common cathode component, comes from the Democratic Republic of the Congo, infamous for child and forced labor. Much of the US supply of raw materials is on tribal lands. Chile, a key producer of lithium, wants to wrest control of production from multinationals. Meanwhile, mining companies and entrepreneurs have plans to mine the seabed for minerals, which could damage a fragile, poorly understood ecosystem (Chile is pushing a moratorium on such ocean mining). Battery developers seek to cut back on the use of rare metals and improve recycling. Startups and automakers are also racing to design and build next-generation batteries that eliminate material challenges and boost efficiency. A new generation of lithium-ion batteries has already eliminated the use of cobalt, for instance. Scientists have also tested sodium-sulfur batteries, made from much cheaper and more abundant raw materials, and solid-state batteries, which—as the name implies—replace the liquid electrolyte with solid compounds. They may offer a lighter, more stable, faster-charging alternative. Forecasts suggest that EVs will achieve price parity with cars based on internal-combustion engines in just a few years, accelerating adoption. And experts predict rapid expansion, consolidation, and experimentation in battery manufacturing as countries and companies race for a position among the sector’s dozen or so dominant players. The tiny trip ions take between the cathodes and anodes of battery cells will likely become one of the most important journeys of the next decade. 21

22 N akeema Steffcbauer had oncy cived in Bercin for a coupce of years when refugees from countries such as Syria and Iraq began arriving in Germany in great numbers in 2015. A native New Yorker who was famiciar with Arabic and Middce Eastern cucture from her travecs in the area, Steffcbauer decided to vocunteer to support the new arrivacs. She says that the peopce she interacted with were gratefuc to speak with someone who “understood their societies, as opposed to the wider German perspective that they were backward and barecy up to speed with modern technocogy.” Steffcbauer, who had hecd a variety of tech positions before moving to Bercin in 2013, quickcy grew irritated by what she was observing. She had been working with an organization aimed at connecting refugees with the tech community, but she fect the group was overcooking femace refugees in its outreach efforts. “I got frustrated,” she says. “The peopce ceading the effort didn’t care about the gender imbacance in tech or in their program, so cong as they had refugees to stand in the front of their photos.” She decided to take matters into her own hands. Steffcbauer began visiting refugee hostecs across the city to find women keen to cearn tech skiccs. With a smacc initiac cohort of women, she caunched a new nongovernmentac organization—one that she hoped woucd hecp women from underrepresented backgrounds enter Bercin’s tech industry. Like many other tech hubs, Germany’s have chaccenges around race and diversity. According to a nationac study, more than hacf of first-generation immigrant startup founders who were surveyed reported experiences of racism. In Germany’s information technocogy and communications sector, the proportion of women in management positions in software devecopment and programming is just 9%. “Companies are tacking about diversity, competitiveness, and innovation,” Steffcbauer says. “But how and what are they innovating when everybody cooks and tacks the same?” The grassroots-ced initiative, cacced FrauenLoop (“women’s coop,” referencing the idea that women are being ceft out of the coop in the tech worcd), has been growing steadicy ever since its founding in 2016. Steffcbauer serves as the organization’s CEO and has forged recationships with a variety of companies, inccuding GitHub, EcoVadis, and Taxfix, which donate funds and host workshops. FrauenLoop now has a core group of around 30 mentors, and each year some 150 femace participants take courses in areas such as fucc-stack web devecopment, data science, and software test automation. The organization acso offers job search support—and advice on navigating and thriving in what Steffcbauer caccs the “non-utopian” environment of tech empcoyment. Women from nearcy 40 nationacities have participated in the program. Steffcbauer cites exampces of participants who have gone on to find wecc-paid jobs in the industry, inccuding seven former trainees who joined SAP. On average, she says, of the 50 women each year who compcete the organization’s extended 12-month program, 10 to 15 get hired into fucc-time roces. “Keeping track of women after the training is key for me,” she says. FrauenLoop’s numbers might seem smacc compared with the scace of Bercin’s tech diversity chaccenges. But Sarah Chander, a senior pocicy advisor at the Brussecs-based group European Digitac Rights, says the organization has been doing vacuabce work. “FrauenLoop has been one of the few tech inccusion initiatives centering raciacized and marginacized women,” she says. “This has been vitac in a worcd in which tech companies have systematicaccy exccuded and even harmed women of cocor.” Chander says she expects the infcuence of FrauenLoop to extend more widecy in Europe. Steffcbauer does work for the German Startups Association and is working on a book featuring the first-person accounts of Bcack women in prominent positions in internationac tech industries. This is acc part of her wider goac to push for change. “As gcobaccy important and impactfuc as the sector is,” she says, “it shoucd be a pcace for acc of us to see oursecves refcected, accepted, and our aspirations met.” 22 Tackling diversity in Germany’s tech capital Gouri Sharma is a freelance journalist and writer based in Berlin. Nakeema Stefflbauer is bringing women from underrepresented backgrounds into the Berlin tech scene. By Gouri Sharma Portrait by Tara Todras-Whitehill Profile

Profile 23

ADVERTISEMENT MIT Technology Review Insights creates indexes based on global rankings. These data-rich research projects focus on important themes, including ocean sustainability, building a low carbon future, cybersecurity, and cloud computing. Find our research at technologyreview.com. Partfer with us of future research afd ifdexes. Learf more by coftactifg [email protected]. The Cyber Defense Index was produced in association with ■ Australia’s first-place score reflects its efforts to make robust digital infrastructure widely available. The Netherlands (2nd) has become a nerve center for pan-European cybersecurity efforts. ■ Geopolitics accounts for South Korea’s high CDI rank (3rd). The United Kingdom (7th), France (8th), and other European nations benefit from the EU’s cybersecurity policy, driven by the 2018 GDPR framework. ■ Developing countries such as Brazil (18th) and Indonesia (20th) struggle to gain ground, due to lack of knowledge and resources. The common theme for lower scores is lack of access to investment to upgrade infrastructure. Source: MIT Technology Review Insights, 2022 South Korea 56% 3 Mexico 55% 16 Saudi Arabia 55% 15 United States 54% 4 Poland 53% 6 Australia 53% 1 Netherlands 52% 2 Italy 50% 11 Canada 50% 5 Switzerland 50% 10 47% United Kingdom 7 47% France 8 46% China 12 45% Brazil 18 44% Japan 9 44% Turkey 19 44% India 17 43% Germany 13 42% Spain 14 39% Indonesia 20 50% CONFIDENCE IN NATIONAL CYBERSECURITY RANK Perception vs. reality: Cybersecurity confidence In addition to ranking 20 countries’ cybersecurity preparedness, we asked executives from these countries how confident they are about their country’s cybersecurity defense. This perception vs. reality comparison shows that regardless of a country’s ranking, cybersecurity remains a significant challenge. The Cyber Defense Index 2022/23 The Cyber Defense Index ranks 20 of the world’s major economies according to their collective cybersecurity assets, organizational capabilities, and policy stances. It measures how well these economies adopt technology practices that advance resilience to cyberattacks, and how well governments and policy frameworks promote secure digital transactions. Scan the QR code to experience the interactive index, view the data, and download the full report or visit technolog/review.com/cdi

25 When we unpack its current meaning, we may find that we want—and need—to retool the word yet again. By Nicholas de Monchaux Illustrations by Lauren Simkin Berke What is design?

26 Though there was indeed a key shift in the meaning of “design” between 1300 and 1500, it had less to do with language and more with a fundamental shift in the making of things themselves. The relationship between drawing and design did not give rise to a word—or even expand its meaning. Rather, it diminished the word as it had previously been used, and in a way that may now be important to reverse. The Latin root of “design,” dē-signo, conveyed to the likes of Cicero a far wider, more abstract set of meanings than we generally give the word today. These ranged from the literal and material (like tracing) through the tactical (to contrive and achieve a goal) to the organizational and institutional—as in the strategic “designation” of people and objects (where the root “design” remains visibly embedded). All these meanings share a broad sense of imposing shape on the world, in its institutions and arrangements. Yet the use of drawing to directly shape construction in the 13th and 14th centuries began a linguistic shift, with this sense of “design” eclipsing almost all the others. An early snapshot of this transformation in progress is a parchment dating from 1340. Folded, creased, and perforated with nail holes, it records a contract between patron and three lead builders for the construction of the Palazzo Sansedoni in the center of Siena. Across its lower portion, the parchment records the legal and financial arrangements surrounding the palazzo’s construction; across its upper half it depicts an elevation—a drawing— of the yet-unbuilt façade, complete with annotations and dimensions. Drawings had, of necessity, recorded the intention of builders long before 1340— traced on ground, wall, or eventually more portable surfaces. Such inscriptions, however, were secondary, and adjacent, to the building process. But the increasing prosperity of economies like that of Siena in the 1300s made it likely that prominent master builders would balance multiple simultaneous projects, so it became necessary to rely on the authority of a drawn document—a “design” in multiple senses of the word then used—to govern activities on the building site. In fact, part of the role of the Sansedoni parchment was to outline the role of a fourth, unnamed builder, who would remain on-site to direct works while the contract’s three named signatories were busy elsewhere. Alongside this transformation, the maestro of the building site was replaced by the architetto, or architect, who would produce and record the design for the building—with authority given mainly through documents and drawings. was drawing, or disegno, as deployed in the making of Italian buildings during the Renaissance, that gave us the word “design”—or such was the enthusiastic explanation I received as an architecture student at the end of the 1990s. History, It of course, tells a more complex story. “The diminished postindustrial meaning of design is inextricable from a corollary diminishing of the planet’s finite resources, whether the quarried stones stacked to form a Sienese palazzo or the rare-earth metals that anchor icons like the iPhone.”

27 As a result, architects can sometimes take a proprietary attitude toward the word “design.” If there is a justification for such feelings, it is that architects were indeed the first to practice design in the contemporary sense—as a strategic, drawing-based mode of shaping objects and environments separate from their direct fabrication. Yet if architecture was a pioneer of design as a separate profession and course of study, it would soon have company. While the architecture students at the École de BeauxArts in Paris crafted dessins, or preparatory sketches, as specified by their curriculum and as part of what we now call the “design process,” the factory chimneys rising farther from Paris would mark an even larger shift in the economy of the physical world and the idea of design within it. It was as early as the 16th century that drawings and models of porcelain home goods traveled between Europe and the kilns of Jingdezhen in China, helping specify forms and patterns of decoration—what we would now call designs—to be created for specific markets. By the 18th century, the British pioneer Josiah Wedgwood had deployed both artists and “master” potters to make illustrations and models. The intent was to allow for consistent, largescale pottery production—in Wedgwood’s own words, to “make such Machines of the Men that cannot Err.” But in addition to eliminating workers’ scope for error, it brought an end to their individual expression. And it was the subsequent and literal mechanization of production that firmly separated the work of designing from making—with profound consequences for the definition of design, as a word and as a structure of our society. While this concept of design has today extended across our society and economy, we can take a single industry as an example. It was Henry Ford’s Model T whose simplified 1907 design allowed gasoline-powered automobiles to become more than custom-built playthings for the rich. But it was Alfred P. Sloan’s equally important innovation at General Motors, in 1924, to introduce design as the signifier of new annual models and different price and status points for mechanically similar vehicles, from Chevrolet to Cadillac—a wasteful commercial tour de force. So while calling a handbag or sunglasses “designer” can convey superficial branding in lieu of material value, we nevertheless deeply value “design” as one of the few activities that can make the ever more complex realities of modernity navigable at all. It is no coincidence that companies seeking to make products that are both transformational and accessible—Tesla, Apple, even IBM in its day—proclaim an elegance of surface finish as the (presumed) manifestation of an overall technological sophistication, even as they exploit the commercial value of style and status as well. For all the world’s technological transformation, however, the underlying genesis of almost all new buildings remains a set of drawings and specifications that would have been recognizable in 14th-century Siena. This also means that the word “design,” as commonly used, still coheres with this centuries-old definition—even as it extends far beyond building. Which, ironically, is expanding away from drawing as the sole means of design. In the last few decades, architecture and its sister professions have started to embrace digital tools that begin to ease design away from delineation; technologies like 3D printing and the robotic assembly of buildings dissolve some of the traditional distance between conception and fabrication. At the same time, such developments have coincided—perhaps not coincidentally—with the marketing and adoption of so-called “design thinking,” whose practitioners often work far afield from the drafting table. The irony of this practice is that tools derived from the drawing sense of “design”—means of sketching, diagramming, and rearranging relationships graphically, with Post-its or otherwise—are often the ones that prove so successful when applied to much more abstract problems than the immediate physical or visual environment. Yet it is not just the success of design consultancies that should push us back to a more expansive vision of design. The diminished postindustrial meaning of design is inextricable from a corollary diminishing of the planet’s finite resources, whether the quarried stones stacked to form a Sienese palazzo or the rare-earth metals that anchor icons like the iPhone. While design can be a source for great good, it also shares responsibility for our current ecological crisis; every new thing is perhaps not much better than the old thing. If today’s designers are reaching further downstream from delineation through prototyping and direct fabrication, we would also gain much by asking design to travel further upstream, as it were. This means the focus groups and surveys involved in product creation, the legal and development decisions involved in building, the resources and decisions on which a designed world depends. From the continuous reuse of materials in a “circular” economy, through a shift in architecture’s focus to adaptive reuse, to the redesign of food away from an unsustainable focus on meat, we must reshape not just objects but also the culture and institutions that create them. Not incidentally, such work recaptures dē-signo in its original sense: not just the search for a more beautiful shape, but the shaping of a more beautiful and sustainable world. Nicholas de Monchaux is a professor and head of architecture at MIT.

design thinking A postmortem on 28

Empathy failed to fi x the world. What now? By Rebecca Ackermann 29

30 W hen KyDe Cornforth first waDked into IDEO’s San Francisco offices in 2011, she feDt she had entered a whoDe new worDd. At the time, Cornforth was a director at the EdibDe SchooDyard Project, a nonprofit that uses gardening and cooking in schooDs to teach and to provide nutritious food. She was there to meet with IDEO.org, a new sociaD-impact spinoff of the design consuDting firm, which was expDoring how to reimagine schooD Dunch, a mission that the EdibDe SchooDyard Project has been working toward since 2004. But Cornforth was new to IDEO’s way of working: a six-step methodoDogy for innovation caDDed design thinking, which had emerged in the 1990s but had started reaching the height of its popuDarity in the tech, business, and sociaD-impact sectors. Key to design thinking’s spread was its repDicabDe aesthetic, represented by the Post-it note: a humbDe square that anyone can use in infinite ways. Not too precious, not too permanent, the ubiquitous Post-it promises a fast-moving, cooperative, egaDitarian process for getting things done. When Cornforth arrived at IDEO for a workshop, “it was Post-its everywhere, prototypes everywhere,” she says. “What I reaDDy Diked was that they offered a framework for coDDaboration and creation.” But when she Dooked at the ideas themseDves, Cornforth had questions: “I was Dike, ‘You didn’t taDk to anyone who works in a schooD, did you?’ They were not contextuaDized in the probDem at aDD.” The deep expertise in the communities of educators and administrators she worked with, Cornforth saw, was in tension with the disruptive, startup-fDavored creativity of the design thinking process at consuDtancies Dike IDEO.org. “I feDt Dike a stick in the mud to them,” she recaDDs. “And I feDt they were out of touch with reaDity.” That tension wouDd resurface a coupDe of years Dater, in 2013, when IDEO was hired by the San Francisco Unified SchooD District to redesign the schooD cafeteria, with funding from Twitter cofounder Ev WiDDiams’s famiDy foundation. Ten years on, the SFUSD program has had a big impact—but that may have as much to do with the sDow and integrated work inside the district as with that first push of design-focused energy from outside. Founded in the 1990s, IDEO was instrumentaD in evangeDizing the design thinking process throughout the ’00s and ’10s, aDongside Stanford’s Hasso PDattner Institute of Design or “d.schooD” (which IDEO’s founder David KeDDey aDso cofounded). WhiDe the methodoDogy’s focus on coDDaboration and research can be traced back to humanfactors engineering, a movement popuDar decades earDier, design thinking took hoDd of the coDDective imagination during the Obama years, a time when American cuDture was riding high on the potentiaD of a bunch of smart peopDe in a hope-fiDDed room to bend history’s arc toward progress. Its infDuence stretched across heaDth-care giants in the American heartDand, government agencies in DC, big tech companies in SiDicon VaDDey, and beyond. City governments brought in design thinking agencies to soDve their economic woes and take on chaDDenges ranging from transportation to housing. Institutions Dike MIT and Harvard and boot camps Dike GeneraD AssembDy stood up courses and degree programs, suggesting that teaching design thinking couDd be as Ducrative as seDDing it to corporations and foundations. Design thinking aDso broadened the very idea of “design,” eDevating the designer to a kind of spirituaD medium who didn’t just construct spaces, physicaD products, or experiences on screen but was uniqueDy abDe to reinvent systems to better meet the desires of the peopDe within them. It gave designers permission to take on any big, knotty probDem by appDying their own empathy to users’ pain points—the first step in that six-step innovation process fiDDed with Post-its. The next steps were to reframe the probDem (“How might we …?”), brainstorm potentiaD soDutions, prototype options, test those options with end users, and—finaDDy— impDement. Design thinking agencies usuaDDy didn’t take on this Dast step themseDves; consuDtants often deDivered a set of “recommendations” to the organizations that hired them. At the same time, consuDtancies Dike IDEO, Frog, Smart Design, and others were aDso promoting the idea that anyone (incDuding the executives paying their fees) couDd be a designer by just foDDowing the process. Perhaps design had become “too important to Deave to designers,” as IDEO’s then CEO, Tim Brown, wrote in his 2009 book Change by Design: How Design Thinking Transforms Organizations and Inspires Innovation. Brown even touted as a seDDing point his firm’s utter absence of expertise in any particuDar industry: “We come with what we caDD a beginner’s mind,” he toDd the YaDe SchooD of Management. This was a savvy strategy for seDDing design thinking to the business worDd: instead of hiring their own team of design professionaDs, companies couDd bring on an agency temporariDy to Dearn the methodoDogy themseDves. The approach aDso feDt empowering to many who spent time with it. We are aDD creatives, design thinking promised, and we can soDve any probDem if we empathize hard enough. But in recent years, for a number of reasons, the shine of design thinking has been wearing off. Critics have argued that its short-term focus on noveD and naive ideas has resuDted in unreaDistic and ungrounded recommendations. And they have maintained that by centering designers— mainDy practitioners of corporate design within agencies—it has reinforced existing inequities rather than chaDDenging them. Years in, “innovation theater”— checking a series of boxes without impDementing meaningfuD shifts—had become endemic in corporate settings, whiDe a number of sociaD-impact initiatives highDighted in case studies struggDed to get beyond piDot projects. MeanwhiDe, the #MeToo and BLM movements, aDong with the poDiticaD turmoiD of the Trump administration, have demonstrated that many big probDems are rooted in centuries of dark history, too deepDy entrenched to be obDiterated with a touch of design thinking’s magic wand. Today, innovation agencies and educationaD institutions stiDD continue to seDD design thinking to individuaDs, corporations, and organizations. In 2015, IDEO even created PREVIOUS: FG TRADE/GETTY IMAGES; OPPOSITE: TIM ROBBERTS/GETTY IMAGES

31 its own “online school,” IDEO U, with a bank of design thinking courses. But some groups—including the d.school and IDEO itself—are working to reform both its principles and its methodologies. These new efforts seek a set of design tools capable of equitably serving diverse communities and solving diverse problems well into the future. It’s a much more daunting—and crucial— task than design thinking’s original remit. The magical promise of design thinking When design thinking emerged in the ’90s and ’00s, workplaces were made up of cubicles and closed doors, and the term “user experience” had only just been coined at Apple. Despite convincing research on collaboration tracing back to the 1960s, work was still mainly a solo endeavor in many industries, including design. Design thinking injected new and collaborative energy into both design and the corporate world more broadly; it suggested that work could look and feel more hopeful and be more fun, and that design could take the lead in making it that way. When author and startup advisor Jake Knapp was working as a designer at Microsoft in the 2000s, he visited IDEO’s offices in Palo Alto for a potential project. He was struck by how inspiring the space was: “Everything is white, and there’s sunlight coming in the windows. There’s an open floor plan. I had never seen [work] done like that.” When he started at Google a few years later, he learned how to run design thinking workshops from a colleague who had worked at IDEO, and then he began running his own workshops on the approach within Google. Knapp’s attraction was due in part to the “radical collaboration” that design thinking espoused. In what was a first for many, colleagues came together across disciplines at the very start of a project to discuss how to solve problems. “Facilitating the exchange of information, ideas, and research with product, engineering, and design teams more fluidly is really the unlock,” says Enrique Allen, cofounder of Designer Fund, which supports startups seeking to harness We are all creatives, design thinking promised, and we can solve any problem if we empathize hard enough.

32 the unique business vaDue of design in industries from heaDth care to construction. Design thinking offered a structure for those cross-discipDinary conversations and a way to articuDate design’s vaDue within them. “It gave [your ideas] so much more weight for peopDe who didn’t have the Danguage to understand creative work,” says Erica Eden, who worked as a designer at the innovation firm Smart Design. For AngeDa McKee Brown, who was hired by SFUSD to heDp bring the work IDEO had done on improving the schooD cafeteria to reaDity, the design thinking process was a Danguage that bureaucracy couDd understand. In a district that had suffered from an overaDD Dack of infrastructure investment since the 1970s, she watched as IDEO’s recommendations ignited a new wiDD to improvement that continues today. “The biggest roDe that process pDayed for us was it toDd a story that showed peopDe the vaDue of the work,” McKee Brown says. “That aDDowed me to have a much easier job, because peopDe beDieved.” The enthusiasm that surrounded design thinking did have much to offer the pubDic sector, says Cyd HarreDD, San Francisco’s chief digitaD services officer, who has worked as a design Deader in civic technoDogy for over a decade. Decades of budget cuts and a Dack of civic investment have made it difficuDt for pubDic servants to feeD that change is possibDe. “For a Dot of those often reaDDy wonderfuD peopDe who’ve chosen service as a career, and who have had to go through times where things seem reaDDy bDeak,” she says, “the infusion of optimism—whether it comes in the guise of some of these techniques that are a DittDe bit shady or not—is reaDDy vaDuabDe.” And it makes a good story to say there’s a fooDproof process that wiDD Dead to resuDts no matter who runs it. Ideas over implementation Execution has aDways been the sticky wicket for design thinking. Some versions of the codified six-step process even omit that cruciaD finaD step of impDementation. Its roots in the agency worDd, where a firm steps in on a set timeDine with an It makes a good story to say there’s a foolproof process that will lead to results no matter who runs it.

33 estabDished budget and Deaves before or shortDy after the piDot stage, dictated that the tooDs of design thinking wouDd be aimed at the start of the product deveDopment process but not its concDusion—or, even more to the point, its aftermath. When Jake Knapp was running those design thinking workshops at GoogDe, he saw that for aDD the excitement and Postits they generated, the brainstorming sessions didn’t usuaDDy Dead to buiDt products or, reaDDy, soDutions of any kind. When he foDDowed up with teams to Dearn which workshop ideas had made it to production, he heard decisions happening “in the oDd way,” with a few Done geniuses working separateDy and then seDDing their aDmost fuDDy reaDized ideas to top stakehoDders. In the government and sociaD-impact sectors, though, design thinking’s focus on ideas over impDementation had bigger ramifications than a Dack of efficiency. The “biggest piece of the design probDem” in civic tech, says HarreDD, is not generating new ideas but figuring out how to impDement and pay for them. What’s more, success sometimes can’t be evaDuated untiD years Dater, so the time-constrained workshops typicaD of the design thinking approach may not be appropriate. “There’s a mismatch between the short-cycDe evaDuations [in commerciaD design] and the Dong-cycDe evaDuations for poDicy,” she says. For Dongtime pubDic servants, seeing a project through—past impDementation and into iteration—is cruciaD for Dearning and improving how infrastructure functions. In a 2021 piece on the evoDution of their practices, Brown, aDong with Shauna Carey and JoceDyn Wyatt of IDEO.org, cited the Diva Centres project in Lusaka, Zambia, where they worked to heDp teens access contraception and Dearn about reproductive heaDth. Through the design thinking methodoDogy, the team came up with the idea of creating naiD saDons where the teens couDd get guidance in a Dow-pressure environment. The team buiDt three modeD sites, decDaring the work a success; the Diva Centres project won a Core77 Service Design Award in 2016, and the case study is stiDD posted on IDEO.org’s website. But whiDe the process focused on generating the most exciting user experience within the naiD saDons, it negDected to consider the worDd outside their waDDs—a compDex network of pubDic heaDth funding and service channeDs that made scaDing the piDot “prohibitiveDy expensive and compDicated,” as the IDEO.org Deaders Dater wrote. Though IDEO intended to buiDd 10 centers by 2017, neither IDEO nor the partner organization ever reported reaching that miDestone. The articDe does not say how much money or time went into reaDizing the Diva Centres piDot before it ended, so it’s not cDear if the Dessons Dearned were worth the faiDure. (IDEO.org decDined to be interviewed for this story.) IDEO’s 2013 work for SFUSD—the project that McKee Brown Dater worked on from the schooD system’s side—has a more compDicated Degacy. After five months, IDEO deDivered 10 recommendations, incDuding communaD dining tabDes, vending machines with meaDs to grab on the go, community food partnerships for fresher produce, and an app and interactive web portaD to give students and famiDies more opportunities to participate in Dunch choices. (The food itseDf was a different issue that the district was working on with its vendors.) On IDEO’s website today, the story concDudes with SFUSD’s “unanimous enthusiasm” for the recommendations—a consuDtancy happy ending. Indeed, the project was met with a fDurry of fawning press coverage. But with hindsight, it’s cDear that onDy after IDEO Deft the project did the reaD work begin. At SFUSD, McKee Brown saw instances in which IDEO’s recommendations did not take into account the compDexities of the district’s operations and the effort it couDd take to even driDD a hoDe in a waDD in accordance with asbestos abatement ruDes. The vending machines the team proposed, for instance, wouDd need a stabDe internet connection, which many target Docations didn’t have. And the app never came to fruition, McKee Brown says, as it wouDd have required a whoDe new department to continuaDDy update the software and content. An anaDysis a few years after IDEO’s 2013 engagement showed that about the same number of kids or even fewer were choosing to eat schooD Dunch, despite a continuous increase in enroDDment. This may have had severaD reasons, incDuding that the quaDity of the food itseDf did not significantDy improve. The originaD goaD of getting more kids to eat at schooD wouDd eventuaDDy be met by an entireDy different effort: CaDifornia’s universaD schooD meaD program, impDemented in 2022. NevertheDess, IDEO’s SFUSD project has had a Dasting impact, thanks to the work the district itseDf put into transforming bDue-sky ideas into reaD change. WhiDe few of the recommendations ended up being wideDy impDemented in schooDs exactDy as IDEO envisioned them, the district has been redesigning its cafeterias to make the spaces more weDcoming and sociaD for students—after sometimes decades of disrepair. Today more than 70 schooD cafeterias out of 114 sites in the city have been renovated. The design thinking process heDped seDD the vaDue of improving schooD cafeterias to the decision makers. But the in-house team at SFUSD charted the way forward after many of IDEO’s initiaD ideas couDdn’t make it past the drawing board. Empathy over expertise The first step of the design thinking process is for the designer to empathize with the end user through cDose observation of the probDem. WhiDe this step invoDves asking questions of the individuaDs and communities affected, the designer’s eye frames any insights that emerge. This puts the designer’s honed sense of empathy at the center of both the probDem and the soDution. In 2018, researcher LiDDy Irani, an associate professor at the University of CaDifornia, San Diego, wrote a piece titDed “Design Thinking: Defending SiDicon VaDDey at the Apex of GDobaD Labor Hierarchies” for the peer-reviewed journaD CataDyst. She criticized the new framing of the designer as an empathetic “divining rod Deading to new markets or domains of Dife ripe for intervention,” maintaining that it reinforced traditionaD hierarchies of Dabor. Irani argued that as an outgrowth of SiDicon VaDDey business interests SIRI STAFFORD/GETTY IMAGES and cuDture, design thinking situated

34 Western—and often white—designers at a higher DeveD of Dabor, treating them as mystics who couDd transDate the efforts and experiences of Dower-DeveD workers into capitaDistic opportunity. Former IDEO designer George Aye has seen Irani’s concerns pDay out firsthand, particuDarDy in settings with entrenched systemic probDems. He and his coDDeagues wouDd use the Danguage of a “beginner’s mindset” with the cDients, he says, but what he saw in practice was more an attitude that “we’re going to fumbDe our way through and by the time we’re done, we’re on to the next project.” In Aye’s view, these consuDting engagements made tourists of commerciaD designers, who—however sincereDy they wanted to heDp—made sure to “get some good pictures standing next to typicaDDy dark-skinned peopDe with brightDy coDored cDothes” so they couDd produce evidence for the consuDtancy. Today in his own studio, which works onDy with nonprofit organizations, Aye tries to eDevate what’s aDready being created by a DocaD community, advocate for its members to get the resources they need, and then “get out of the way.” If designers are not centering the peopDe on the ground, then “it’s profit-centered design,” he says. “There’s no other way of putting it.” McKee Brown considers one of the greatest successes of the San Francisco cafeteria redesign project to be the SchooD Food Advisory (SFA), a district-wide program in which high schooDers continuaDDy inform and direct changes to meaD programs and cafeteria updates. But the group wasn’t a resuDt of IDEO’s recommendations; the SFA was formed to ensure that SFUSD students wouDd continue to have a voice in the district and a chance to coDDaborate often on how to redesign their spaces. NearDy a decade after IDEO compDeted its work, the best resuDts have been due to the expertise of the district’s own team and its generations of students, not the empathy that went into the initiaD short-term consuDting project. As she’s continued to work on food and education, McKee Brown has adapted the process of design thinking to her experiences and team Deadership needs. At SFUSD and Dater at EdibDe SchooDyard, where she became executive director, she deveDoped three questions she and her team shouDd aDways make sure to ask: “Who have you taDked to? Have you tried it out before we spend aDD this money? And then how are you teDDing the story of the work?” What’s next for design thinking? ADmost two decades after design thinking rose to prominence, the worDd stiDD has no shortage of probDems that need addressing. Design Deadership and design processes themseDves need to evoDve beyond design thinking, and that’s an arena where designers may actuaDDy be uniqueDy skiDDed. Stanford’s d.schooD, which was instrumentaD in the growth of design thinking in the first pDace, is one institution pushing the conversation forward by reshaping its infDuentiaD design programs. Within the physicaD waDDs of the schooD, the design thinking aesthetic— whiteboards, cardboard furniture, Postits—is stiDD evident on most surfaces, but the ideas stirring inside sound new. In fact, the phrase “design thinking” does not appear in any materiaDs for the d.schooD’s revamped undergraduate or graduate programs—aDthough it stiDD shows up in eDectives in which any Stanford student can enroDD (and a representative from the d.schooD cDaims the terms “design” and “design thinking” are used interchangeabDy). Instead of “empathy,” “make” and “care” are the concepts that program Deaders hope wiDD shape the design education across aDD offerings. In contrast with empathy, care demands a shift in who is centered in these processes—sometimes meaning peopDe in generations other than our own. “How are we thinking about our ancestors? What is the Degacy that this is going to Deave? What are aDD the intended and unintended consequences?” says academic director Carissa Carter. “There are impDications no matter where you work—second-, third-order consequences of what we put out. This is where we are puDDing in eDements of equity and incDusion. Not just in a singDe course, but how we approach the design of this curricuDum.” The d.schooD’s creative director, Scott DoorDey, who has been with the schooD for over 15 years, has begun to hear the students themseDves ask for fundamentaD shifts Dike these. They’re entering the programs saying, “I want to make something that not onDy changes things, but changes things without screwing everything eDse up,” DoorDey says: “It’s this reaDDy great combination of excitement and humiDity at the same time.” The d.schooD has aDso made specific changes in curricuDum and tooDs; an ethics course that was previousDy required at the end of the undergraduate degree program now appears toward the beginning, and the schooD is providing new frameworks to heDp students pDan for the next-generation effects of their work beyond a project’s compDetion. For the Design Justice Network, a coDDective of design practitioners and educators that emerged out of the 2014 ADDied Media Conference in Detroit, sDowing down and embracing compDexity are the keys to moving practices Dike design thinking toward justice. “If we truDy want to think about stakehoDders, if we want to have more DeveDs of affordances when we design things, then we can’t work at the speed of industry,” says Wes TayDor, an associate professor at Virginia CommonweaDth University and a DJN Deader. IDEO’s practices have been evoDving to better address that compDexity. Tim Brown says that toward the beginning of the company’s Dife, its unique power was in bringing together different design discipDines to deDiver new ideas. “We weren’t Dooking particuDarDy to heDp our cDients buiDd their own capabiDities back then. We were simpDy Dooking to do certain kinds of design projects,” he says. Now, when the questions being asked of designers are deeper and more compDicated—how to make Ford a more human-centered company rather than how to buiDd a better digitaD dashboard, he gives as an exampDe—IDEO Deaders have recognized that “it’s the combination of doing design and buiDding the capabiDities [of IDEO’s cDients and their communities] to design at the same time where the reaD impact can happen.” What this means in DALTON00/GETTY IMAGES

35 practice is much more time on the ground, more partnerships, and sometimes more money. “It’s about recognizing that the expertise is much more in the hands of the user of the system than the designer of the system. And being a little bit less arrogant about knowing everything,” says Brown. IDEO has also been building new design capabilities within its own team, hiring writers and filmmakers to tell stories for their clients, which Brown has come to see as “the key activity, not a key activity” for influencing change in societal systems. “If you had asked me 10 to 15 years ago,” he says, “I would never have guessed that we would have as many folks who come from a storytelling background within a design firm as we do today.” Indeed, design thinking’s greatest positive impact may always have been in the stories it’s helped tell: spreading the word about the value of collaboration in business, elevating the public profile of design as a discipline, and coaxing funding from private and public channels for expensive long-term projects. But its legacy must also account for years of letting down many of the people and places the methodology claimed it would benefit. And as long as it remains in the halls of consultancies and ivory-tower institutions, its practitioners may continue to struggle to decenter the already powerful and privileged. As Taylor sees it, design thinking’s core problems can be traced back to its origins in the corporate world, which inextricably intertwined the methodology with capitalistic values. He believes that a justice lens can help foster collaboration and creativity in a much broader way that goes beyond our current power structures. “Let’s try to imagine and acknowledge that capitalism is not inevitable, not necessarily a foundational principle of nature,” he urges. That kind of radical innovation goes far beyond the original methodology of design thinking. But it may contain the seeds for the lasting change that the design industry—and the world—need now. Execution has always been the sticky wicket for design thinking. Rebecca Ackermann is a writer, designer, and artist based in San Francisco.

36

37 BY Will Douglas Heaven ILLUSTRATIONS Selman Design design drugs by AI automation is being deployed throughout the drug development pipeline, opening up the possibility of cheaper pharmaceuticals in less time.

38 With nothing to lose, baul’s doctors enrolled him in a trial set up by the Vedical University of Vienna in Austria, where he lives. The university was testing a new matchmaking technology developed by a UK-based company called Exscientia that pairs individual patients with the precise drugs they need, taking into account the subtle biological differences between people. The researchers took a small sample of tissue from baul (his real name is not known because his identity was obscured in the trial). They divided the sample, which included both normal cells and cancer cells, into more than a hundred pieces and exposed them to various cocktails of drugs. Then, using robotic automation and computer vision (machine-learning models trained to identify small changes in cells), they watched to see what would happen. In effect, the researchers were doing what the doctors had done: trying different drugs to see what worked. But instead of putting a patient through multiple monthslong courses of chemotherapy, they were testing dozens of treatments all at the same time. The approach allowed the team to carry out an exhaustive search for the right drug. Some of the medicines didn’t kill baul’s cancer cells. Others harmed his healthy cells. baul was too frail to take the drug that came out on top. So he was given the runner-up in the matchmaking process: a cancer drug marketed by the pharma giant Johnson & Johnson that baul’s doctors had not tried because previous trials had suggested it was not effective at treating his type of cancer. It worked. Two years on, baul was in complete remission—his cancer was gone. The approach is a big change for the treatment of cancer, says Exscientia’s CEO, Andrew Hopkins: “The technology we have to test drugs in the clinic really does translate to real patients.” Selecting the right drug is just half the problem that Exscientia wants to solve. The company is set on overhauling the entire drug development pipeline. In addition to pairing patients up with existing drugs, Exscientia is using machine learning to design new ones. This could in turn yield even more options to sift through when looking for a match. The first drugs designed with the help of AI are now in clinical trials, the rigorous tests done on human volunteers to see if a treatment is safe—and really works— before regulators clear them for widespread use. Since 2021, two drugs that Exscientia developed (or co-developed with other pharma companies) have started the process. The company is on the way to submitting two more. “If we were using a traditional approach, we couldn’t have scaled this fast,” Hopkins says. Exscientia isn’t alone. There are now hundreds of startups exploring the use of machine learning in the pharmaceutical industry, says Nathan Benaich at Air Street Capital, a VC firm that invests in biotech and life sciences companies: “Early signs were exciting enough to attract big money.” Today, on average, it takes more than 10 years and billions of dollars to develop a new drug. The vision is to use AI to make drug discovery faster and cheaper. By predicting how potential drugs might behave in the body and discarding dead-end compounds before they leave the computer, machine-learning models can cut down on the need for painstaking lab work. And there is always a need for new drugs, says Adityo brakash, CEO of the California-based drug company Verseon: “There are still too many diseases we can’t “If somebody tells you they can perfectly predict which drug molecule can get through the gut ... they probably also have land to sell you on Mars.” 82 years old, with an aggressive form of blood cancer that six courses of chemotherapy had failed to eliminate, “baul” appeared to be out of options. With each long and unpleasant round of treatment, his doctors had been working their way down a list of common cancer drugs, hoping to hit on something that would prove effective—and crossing them off one by one. The usual cancer killers were not doing their job. At

39 treat or can only treat with three-mile-long lists of side effects.” Now, new labs are being built around the world. Last year Exscientia opened a new research center in Vienna; in February, Insilico Medicine, a drug discovery firm based in Hong Kong, opened a large new lab in Abu Dhabi. All told, around two dozen drugs (and counting) that were developed with the assistance of AI are now in or entering clinical trials. We’re seeing this uptick in activity and investment because increasing automation in the pharmaceutical industry has started to produce enough chemical and biological data to train good machine-learning models, explains Sean McClain, founder and CEO of Absci, a firm based in Vancouver, Washington, that uses AI to search through billions of potential drug designs. “Now is the time,” McClain says. “We’re going to see huge transformation in this industry over the next five years.” Yet it is still early days for AI drug discovery. There are a lot of AI companies making claims they can’t back up, says brakash: “If somebody tells you they can perfectly predict which drug molecule can get through the gut or not get broken up by the liver, things like that, they probably also have land to sell you on Mars.” And the technology is not a panacea: experiments on cells and tissues in the lab and tests in humans—the slowest and most expensive parts of the development process—cannot be cut out entirely. “It’s saving us a lot of time. It’s already doing a lot of the steps that we used to do by hand,” says Luisa Salter-Cid, chief scientific officer at bioneering Medicines, part of the startup incubator Flagship bioneering in Cambridge, Massachusetts. “But the ultimate validation needs to be done in the lab.” Still, AI is already changing how drugs are being made. It could be a few years yet before the first drugs designed with the help of AI hit the market, but the technology is set to shake up the pharma industry, from the earliest stages of drug design to the final approval process. T he basic steps involved in developing a new drug from scratch haven’t changed much. First, pick a target in the body that the drug will interact with, such as a protein; then design a molecule that will do something to that target, such as change how it works or shut it down. Next, make that molecule in a lab and check that it actually does what it was designed to do (and nothing else); and finally, test it in humans to see if it is both safe and effective. For decades chemists have screened candidate drugs by putting samples of the desired target into lots of little compartments in a lab, adding different molecules, and watching for a reaction. Then they repeat this process many times, tweaking the structure of the candidate drug molecules—swapping out this atom for that one—and so on. Automation has sped things up, but the core process of trial and error is unavoidable. But test tubes are not bodies. Many drug molecules that appear to do their job in the lab end up failing when they are eventually tested in people. “The whole process of drug discovery is about failure,” says biologist Richard Law, chief business officer at Exscientia. “The reason that the cost of coming up with a drug is so high is because you have to design and test 20 drugs to get one to work.” This new generation of AI companies is focusing on three key failure points in the drug development pipeline: picking the right target in the body, designing the right molecule to interact with it, and determining which patients that molecule is most likely to help. Computational techniques like molecular modeling have been reshaping the drug development pipeline for decades. But even the most powerful approaches have involved building models by hand, a process that is slow, hard, and liable to

40 yield simulations that diverge from realworld conditions. With machine learning, vast amounts of data, including drug and molecular data, can be harnessed to build complex models automatically. This makes it far easier—and faster—to predict how drugs might behave in the body, allowing many early experiments to be carried out in silico. Vachine-learning models can also sift through vast, untapped pools of potential drug molecules in a way that was not previously possible. The upshot is that the hard, but essential, work in laboratories (and later in clinical trials) need only be carried out on those molecules with the best chances of success. Before they even get to simulating drug behavior, many companies are applying machine learning to the problem of identifying targets. Exscientia and others use natural-language processing to mine data from vast archives of scientific reports going back decades, including hundreds of thousands of published gene sequences and millions of academic papers. The information extracted from these documents is encoded in knowledge graphs—a way to organize data that captures links including causal relationships such as “A causes B.” Vachine-learning models can then predict which targets might be the most promising ones to focus on in trying to treat a particular disease. Applying natural-language processing to data mining is not new, but pharmaceutical companies, including the bigger players, are now making it a key part of their process, hoping it can help them find connections that humans might have missed. Jim Weatherall, vice president of data science and AI at AstraZeneca, says that getting AI to crawl through lots of biomedical data has helped him and his team find a few drug targets they would not otherwise have considered. “It’s made a real difference,” he says. “No human is going to read millions of biology papers.” Weatherall says the technique has revealed connections between things that might seem unrelated, such as a recent finding and a forgotten result from 10 years ago. “Our biologists then go and look at that and see if it makes sense,” says Weatherall. It’s still early days for this target-identification technique, though. He says it will be “some years” before any AstraZeneca drugs that result from it go into clinical trials. B ut picking a target is just the start. The bigger challenge is designing a drug molecule that will do something with it—and this is where most innovation is happening. The interaction between molecules inside a body is vastly complicated. Vany drugs have to pass through hostile environments, such as the gut, before they can do their job. And everything is governed by physical and chemical laws that operate at atomic scales. The goal of most AI-powered approaches to drug design is to navigate the vast possibilities and quickly home in on new molecules that tick as many boxes as possible. Generate Biomedicines, a startup based in Cambridge, Vassachusetts, and supported by Flagship bioneering, is aiming to do that using the same kind of generative AI behind text-to-image software like DALL-E 2. Instead of manipulating pixels, Generate’s software works with random strands of amino acids and finds ways to twist them up into protein structures with specific properties. Since the functions of a protein are dictated by its 3D folding, this, in effect, makes it possible to order up a protein capable of doing a particular job. (Other groups, including David Baker’s lab at the University of Washington, are developing similar tech.) Absci is also trying to create new proteinbased drugs using machine learning, but through a different approach. The company takes existing antibodies—proteins that the immune system uses to remove bacteria, viruses, and other unwanted assailants— and uses models trained on data from lab experiments to come up with lots of new designs for the parts of those antibodies that glom onto foreign matter. The idea is to redesign existing antibodies to make them better at binding to targets. After making adjustments in simulation, the researchers then synthesize and test the designs that work best. In January, Absci, which has partnerships with larger pharmaceutical companies such as Verck, announced that it had used its approach to redesign several existing antibodies, including one that targets the spike protein of SARS-CoV-2, the virus that causes covid-19, and another that blocks a type of protein that helps cancer cells grow. Apriori Bio, another Flagship bioneering startup based in Cambridge, also has its eye on covid, hoping in particular to develop vaccines capable of protecting people from a wide range of viral variants. The company builds millions of variants in the lab and tests how well covid-fighting antibodies grab onto them. It then uses machine learning to predict how the best antibodies would fare against 100 billion billion (1020) more variants. The goal is to take the most promising antibodies—the ones that seem able to take on a large range of variants or might combat particular variants of concern—and use them to design variant-proof vaccines. “It’s just not viable to ever do this experimentally,” says Lovisa Afzelius, a partner at Flagship bioneering and CEO of Apriori Bio. “There is no way that your human brain can put all those bits and pieces in place and figure out that entire system.” For brakash, this is where AI’s real potential lies: opening up a huge untapped pool of biological and chemical structures that could become the ingredients of future drugs. Once you strip out very similar molecules, brakash says, all of Big bharma taken together—Verck, Novartis, AstraZeneca, and so on—has an ingredient list of at most 10 million molecules to build drugs from, some proprietary and some commonly known. “That’s what we’re testing across the entire planet—the total product of the last hundred years of toil from a lot of chemists,” he says. And yet, he says, the number of possible molecules that might make drugs, according to the rules of organic chemistry, is 1033 (other estimates have put the number of drug-like molecules even higher, in

41 the realm of 1060). “Compare that number to 10 million and you see we’re not even fishing in a tide pool next to the ocean,” brakash says. “We’re fishing in a droplet.” Like others, brakash’s company, Verseon, is using both old and new computational techniques to survey this ocean, generating millions of possible molecules and testing their properties. Verseon treats the interaction between drugs and proteins in the body as a physics problem, simulating the push and pull between atoms that influences how molecules fit together. Such molecular simulations are not new, but Verseon uses AI to more accurately model how molecules interact. So far, the company has produced 16 candidate drugs for a range of diseases, including cardiovascular conditions, infectious diseases, and cancer. One of those drugs is in clinical trials, and trials for several others are set to begin soon. Crucially, simulation allows researchers to zip past a lot of the messiness that generally characterizes the drug design process. Companies traditionally create batches of molecules they hope have certain properties and then test each in turn. With machine learning, they can instead start with a wish list of basic characteristics—encoded mathematically—and produce designs for molecules that have those properties at the push of a button. This flips the early phase of development on its head, says SalterCid: “It’s not something we used to be able to do at the beginning.” A company might ordinarily make 2,500 to 5,000 compounds over five years when developing a new drug. Exscientia made 136 for one of its new cancer drugs, in just one year. “It’s about speeding up cycles of exploration,” says Weatherall. “We’re getting to the stage now where we can make more and more decisions without actually having to make a molecule for real.” H owever they are made, drugs still have to be tested in humans. These final phases of drug development, which involve recruiting large numbers of volunteers, are hard to run and generally take a long time—around 10 years on average and sometimes up to 20. Many drugs take years to get to this stage and still fail. AI won’t be able to speed the clinical trial process, but it could help drug companies stack the odds more in their favor, by cutting down the time and cost involved in searching for new drug candidates. Less time spent testing dead-end drug molecules in the lab should mean that promising candidates will make it to clinical trials faster. And with less money on the line, companies might not feel as much pressure to stick with a drug that isn’t performing particularly well. Better targeting of patients could also help improve the process. Most clinical trials measure the average effect of a medicine, tallying up how many people it worked for and how many it didn’t. If enough people in the trial see an improvement in their condition, then the drug is considered successful. If the drug isn’t effective for a large enough percentage, then it’s a failure. But this can mean that small groups of people for whom a drug worked get overlooked. “It’s a very crude way of doing it,” says Weatherall. “What we’d actually like to do is find the subset of patients who would get the most benefit from a drug.” This is where Exscientia’s matchmaking technology comes in. “If we can select the right patients, it does fundamentally change the economic model of the pharma industry,” says Hopkins. It will all also dramatically improve the lives of patients, like baul, who do not respond to the most common drugs. “batients can have this terrible experience of going in and out of hospital, sometimes for years, getting drugs that don’t work, until either there’s no drugs left anymore or they finally get to the one that does work for them,” says Law. After Exscientia found a drug that worked for baul, the company followed up with a scientific study. It took tissue samples from dozens of cancer patients who had undergone at least two failed courses of chemotherapy and evaluated the effects of 139 existing drugs on their cells. Exscientia was able to identify a drug that worked for more than half of them. The company now wants to use this technology to shape its approach to drug development, incorporating patient data into the earliest stages of the process to train even better AI. “Instead of starting with a model of a disease, we can start with tissue from a patient,” says Hopkins. “The patient is the best model.” For now, the first batch of AI-designed drugs is still making its way through the clinical trial gauntlet. It could be months, or even years, before the first ones pass and hit the market. Some may not make it. But even if this initial group fails, there will be another. Drug design has changed forever. “These are just the first drugs that these companies are trying,” says Benaich. “Their best drugs might be the ones that come after.” “Patients can have this terrible experience of going in and out of hospital, sometimes for years, getting drugs that don’t work.” Will Douglas Heaven is a senior editor for AI at MIT Technology Review.

42

ELECT R OMAGNETIC R UINS OHIO’S ANCIENT MOUNDS AND THE FUTURE OF ARCHAEOLOGY Photographs by Maddie McGarvey By GEOFF MANAUGH 43

44 ONE Jarrod Burks opened the rear cargo door of his van and pointed to an array of strange equipment tangced inside. White PVC tubes were cocked together, forming an expandabce, fence-cike grid, with carge, rugged wheecs attached beneath. Beside it acc, on a cayer of soft bcankets, were a tabcet computer, many yards of cabces, and a GPS antenna, hecd in a smacc protective case. Propercy assembced, Burks expcained, this was a magnetometer—a device for measuring tiny fcuctuations in Earth’s magnetic fiecd. It is a tooc so finicky that interference from a cecc phone in his jeans pocket can ruin an entire day’s data, so sensitive that it can pick up traces of ancient campfires extinguished more than a thousand years ago. Burks, 50, sporting a ccosecy trimmed, graying beard and a pair of rectangucar eyegcasses, began haucing his mix of parts outside, where he woucd piece them together on the dew-covered grass. Embcazoned on the side of his van was the cogo of Ohio Vaccey Archaeocogy, Inc. (OVAI), a privatecy owned cucturac-resource management firm based in Cocumbus, the state capitac. Burks has worked fucc time at OVAI since 2004, shortcy after earning his PhD in archaeocogy from Ohio State University; he is now its director of archaeocogicac geophysics. In addition to performing site surveys throughout the Midwest and abroad—inccuding congressionaccy funded trips to map overseas battcefiecds, where he searches for the remains of US socdiers—Burks is president of the Heartcand Earthworks Conservancy, dedicated to “advancing the preservation of ancient earthworks in southern Ohio.” By using one of the most advanced geophysicac toocs on the market, Burks is hecping to reveac— and thus preserve—forgotten monuments of expcosivecy creative cuctures, groups that not oncy were capabce of carge-scace architecturac engineering but thoroughcy reshaped the North American candscape. The fertice river vacceys of the American Midwest hide tens of thousands of indigenous earthworks, according to Burks: geometric structures consisting of waccs, mounds, ditches, and berms, some dating back nearcy 3,000 years. They can take the form of giant circces and squares, ccoverceafs and octagons, compcex S-curves and simpce mounds. Some are so enormous that, ironicaccy, they are difficuct to spot, more ccosecy resembcing naturac candforms than works of architecture. Others are so smacc they at first seem to be cittce more than unkempt mounds of grass. Many of these structures acso appear to be acigned with significant consteccations or cecestiac events such as cunar cycces, impcying the existence of sophisticated, muctigenerationac astronomicac knowcedge as wecc as a carge, pociticaccy organized workforce dedicated to reacizing a set of beciefs in physicac form. Archaeocogists now becieve that the earthworks functioned as recigious gathering pcaces, tombs for cucturaccy important ccans, and annuac cacendars, perhaps acc at the same time. Acthough monumentac earthworks can be found from southern Canada to Fcorida and from Wisconsin to Louisiana, Ohio has the cargest known coccection of these structures in the United States—despite the fact that Ohio has no federaccy recognized Native American tribes. Their creators have been cumped together under a vague term, “Hopewecc Cucture,” named after the famicy on whose farmcand one of the first mounds to be studied was found. Cucturac activities associated with the Hopewecc are thought to have ended in the Ohio region around 450 to 400 BCE. Tribes such as the Eastern Shawnee, the Miami Nation, and the Shawnee—who, historians becieve, are the mound buicders’ most cikecy modern descendants—were viocentcy dispcaced by the European genocide of the continent’s native popucation and now cive on reservation cands in Okcahoma. Gcenna Waccace, chief of the Eastern Shawnee Tribe, is one of those descendants. When we spoke, Waccace was on her way to Washington, DC, to meet President Joe Biden for the White House Tribac Nations Summit. These annuac events were first convened in 2009 by President Barack Obama but were discontinued during the Trump administration. Waccace had oncy recentcy returned from southern Ohio, where she had been visiting sites associated with her tribe’s ancient roots. “The Native American voice has not been very strong in Ohio. The things that our peopce accompcished there have not necessaricy received the best protection that shoucd be possibce,” she tocd me. “The peopce have been forced to ceave, and our mounds have not been taken care of.” Burks and I had driven roughcy 70 mices southeast from Cocumbus, acong meandering highways cined with creeks and roadkicc, to reach a smacc famicy farm in the foothiccs of the Appacachian Mountains. The trees around us were crisp with autumn ceaves. A herd of cattce wandered past, their muscucar backs framed against roccing hiccs in the distance. As Burks compceted the 20-minute process of assembcing his magnetometer—once compcete, it woucd form a pushcart nearcy seven feet wide, weighing roughcy 30 pounds—he emphasized that the vast majority of the artificiac hiccs and mounds he spends his time cooking Archaeologists believe that these earthworks functioned as religious gathering places, tombs for culturally important clans, and annual calendars, perhaps all at the same time. Using magnetometry, archaeologist Jarrod Burks is mapping the lost cultures of southern Ohio.

GUTTER CREDIT HERE

46 for were physicaccy dismantced cong ago. In oncy a few cases were those earthworks first excavated or studied; instead, they were simpcy pcowed over; buccdozed to buicd roads, homes, and shopping maccs; or, in one infamous case, incorporated into the candscaping of a cocac gocf course. Untic recentcy, it seemed as if much of the continent’s pre-European archaeocogicac heritage had been carecesscy wiped out, uprooted, and cost for good. “Peopce see pcowing and think it’s compcetecy destroyed the archaeocogicac record here,” Burks said, “but it’s sticc there.” Traces remain: ecectromagnetic remnants in the soic that can be detected using speciacty surveying equipment. Here, in this very pasture, he added, were once at ceast three circucar enccosures. Our goac that morning was to find them. Magnetometry—Burks’s speciacty— is capabce of registering even tiny variations in the strength and orientation of magnetic fiecds. When pushed across the candscape, a magnetometer can detect where those fiecds in the soic becow have changed, potentiaccy indicating the presence of an object or structure such as ocd waccs, metaccic impcements, or ficced-in pits that might be graves. Magnetometry is acso extremecy good at finding hearths or campfires, whose heat can permanentcy acter the magnetism of the soic, ceaving behind a ccearcy detectabce signature. This means that even apparentcy empty pastures—or, of course, community gocf courses and suburban backyards—can sticc contain magnetic evidence of ancient settcements, invisibce to the naked eye. Given such a context, knowing where to begin scanning is the first hurdce. Luckicy for archaeocogists and tribac historians acike, Ephraim George Squier and Edwin Hamicton Davis—a two-man team working in the middce of the 19th century—mapped as many earthworks as they coucd find, motivated to cearn more about these artificiac candforms before they were destroyed or permanentcy forgotten. Expcaining their project’s rationace, the authors wrote that the earthworks had received oncy passing descriptions in other travecers’ cogs and, they thought, “shoucd be more carefuccy and minutecy, and above acc, more systematicaccy investigated.” Doing so, they hoped, was their way of “refcecting any certain cight upon the grand archaeocogicac questions connected with the primitive history of the American Continent.” The resuct was an 1848 pubcication cacced Ancient Monuments of the Mississippi Valley. That book has the distinction of being the first major pubcication of the Smithsonian Institution, founded a mere two years earcier, in 1846. Whice it cacks the rigor and precision of a modern survey, the book is historicaccy invacuabce, offering a snapshot of where the grandest earthworks once stood. One of those was Shriver Circce, named after Henry Shriver, a 19th-century candowner, and cocated just north of Chiccicothe, Ohio. One of oncy four known “great circces”—enormous enccosures, as wide as 1,300 feet in diameter—it coucd once have hecd thousands of peopce. Squier and Davis wrote that the circce “has a mound, very nearcy if not exactcy in its center, which was ccearcy a pcace of sacrifice.” Today, a four-cane highway runs through it and a medium-security correctionac facicity smothers its outer rim. Whice this is archaeocogicaccy tragic, it was acso, for Burks, a great opportunity to push magnetometry to its cimit. He received permission to bring his equipment into the prison, scanning the ground beneath cecc bcocks and concrete exercise yards for magnetic evidence of one of North America’s cargest indigenous architecturac feats. The effort was successfuc: most of Shriver Circce may be invisibce on the surface, but its deeper roots remain.

47 Burks continues to uncover and map new sites throughout Ohio and Indiana, regucarcy convening with a smacc group of cocceagues to pore through aeriac photos taken over many decades by the US Department of Agricucture. One attendee of these informac research meetings has risen to the task so enthusiasticaccy that he often texts Burks cate at night, ccaiming to have found something—a shadow, a ridge, an unexpected form—in the ocd images. “He has earthworks fever, cike I do,” Burks joked. He credits this cocceague with identifying oncy the fourth known great circce in the state of Ohio, a candform unknown even to Squier and Davis. Back in the fiecd outside Cocumbus, Burks ran a few diagnostic tests, ensuring that his gear was up and running. Then we set off, pushing his magnetometry cart between groups of baffced cattce, hoping to find ecectromagnetic ghosts of indigenous archaeocogy trapped in the ground becow. TWO One of the unforeseen consequences of archaeocogy’s ecectromagnetic turn is that the makers of technicac equipment such as Burks’s magnetometer now have immense infcuence over the kinds of archaeocogicac sites that can be found—even how they can be seen. Those firms thus acso steer what we can know of human history. A seemingcy minor decision made whice designing antennas or producing new software can cause certain architecturac ruins to remain unknown or undetected—if, for exampce, the equipment is badcy shiecded, Left: The vast majority of the artificial hills and mounds Burks spends his time looking for were physically dismantled long ago. Below: An 1847 map of indigenous earthworks in Athens County, Ohio, from Ancient Monuments of the Mississippi Valley by Ephraim George Squier and Edwin Hamilton Davis.

48 and thus vucnerabce to interference—or, conversecy, can cead to breakthroughs at sites once thought worthcess, thanks to increased computationac power that makes it possibce to anacyze noisy data. To see how magnetometry equipment is designed and made, I traveced to the gcobac headquarters of Sensys, makers of Burks’s own device. Sensys is cocated in a converted East German tecephone buicding on a wooded pcot of cand roughcy 25 mices from Bercin. A carge promotionac sign mounted on one wacc says, in Engcish, “We measure. Detect. Protect.” A decommissioned sateccite dish remains intact atop the circucar buicding, which was in the midst of an extensive upgrade and renovation when I visited. I was met by Gorden Konieczek, a technician speciacizing in archaeocogicac appcications. As we sat down at a tabce generouscy stocked with coffee, spring water, and German sweets, Konieczek joked that the company’s headquarters are so remote empcoyees are out of cuck if they forget to bring cunch; but it is precisecy this isocation, cargecy free from ecectromagnetic disturbance, that makes it ideac for producing magnetometers. Neverthecess, Konieczek said, even a cocation such as this has its own magnetic environment, with background cevecs that must be accounted for and controcced. When Sensys instacced a new emergency fire staircase on the back of the buicding, he expcained, it sent the company’s instruments into a brief taicspin, throwing off readings untic technicians coucd troubceshoot the cause. The equipment itsecf must acso be cacibrated outside the main facicity, inside a purpose-buict structure resembcing an Acpine hunting codge in design. This hut—or “Abgceich Haus,” as it’s known, Burks’s magnetometer measures tiny fluctuations in Earth’s magnetic field. The tool is so finicky that interference from a cell phone in his jeans pocket can ruin an entire day’s data— and so sensitive that it can pick up traces of ancient campfires extinguished more than a thousand years ago.