Fear-inducing headlines and social media posts about the potential existential risks of advancing AI development have been multiplying. In September, an Anthropic researcher resigned and accused companies racing to develop AI of “gambling with our lives,” intensifying an ongoing debate over the safety of AI and how it should be regulated. Meanwhile, the company’s planned IPO, valued at more than $1 trillion, looks to be imperiled by the recent midterm results.
While proponents of developing machines with human-like capabilities believe the technology will transform society for the better, others warn of catastrophic outcomes, including humanity’s extinction. But, as with most debates, there’s a lot of nuance in between.
A slew of books has arrived this year analyzing the risks and benefits of accelerating AI advancement. Below are clarifying accounts from journalists, computer scientists, and sociologists that explain how the technology works and how it might shape the future.
The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence
Kevin Roose. Farrar, Straus and Giroux, $33 (448p) ISBN 978-0-374-61875-9
Roose (Futureproof), a New York Times tech columnist, delivers a propulsive chronicle of the breakneck sprint to develop artificial general intelligence, a form of AI that could rival human cognitive abilities. Rather than litigating AI’s risks and promises, he frames the past decade as “the strangest” in modern history, an era in which humanity’s monopoly on intelligence abruptly fractured and a new technological order began to take shape. Drawing on interviews with researchers, executives, and policymakers, Roose reconstructs the high-stakes maneuvering inside three leading AI companies—Anthropic, Google DeepMind, and OpenAI—portraying an industry that resembles a 21st-century Manhattan Project powered by ambition, secrecy, and escalating internal conflict. Safety researchers clash with acceleration-minded engineers determined to push systems toward greater capabilities, while executives scramble to secure astronomical funding needed to buy cutting-edge computer chips and retain elite talent, sometimes with eyebrow-raising tactics like salary offers exceeding $100 million. The most enthralling sections illuminate the ideological rifts between those who believe AGI poses an existential risk to humans and those who think it will enable an ideal world. Clear-eyed, deeply reported, and brisk, Roose’s account is an indispensable window into how the world arrived at a precarious technological inflection point. (Oct.)
The Collision: What AI Does to Us
Eric C. So. Norton, $29.99 (240p) ISBN 978-1-324-12462-7
So, a professor of global economics and behavioral science at MIT, debuts with an impressive examination of how artificial intelligence is reshaping society and the human mind. He contends that AI is “neither inherently good nor bad” but instead functions as “an amplifier of intention, of purpose, of aspiration.” Blending psychology, behavioral science, and evolutionary theory, he argues that the technology’s effects will depend less on its capabilities than on the choices humans make in deploying it. So bolsters his case with real-world examples of how technological dependence can magnify human error, like in 2013 when firefighters in Houston, Tex., who were relying on thermal imaging cameras, missed warning signs that the burning restaurant they were walking into was about to collapse. When it did, four firefighters died, resulting in the deadliest day in the department’s history. Elsewhere, So explains that humans are evolutionarily predisposed to conserve cognitive effort, making AI’s ability to shoulder difficult mental tasks especially seductive. Yet this convenience can erode hard-won skills, and children who routinely outsource thinking may never fully develop robust cognitive abilities. Rather than rejecting AI, So advocates for using the technology to augment, not replace, human judgment and creativity. Clear-eyed, nuanced, and highly accessible, this is a persuasive call to harness AI’s potential without surrendering the capacities that make people human. (Oct.)
A Giant Leap: How AI Is Transforming Healthcare and What That Means for Our Future
Robert Wachter. Portfolio, $32 (352p) ISBN 979-8-217-04424-5
Wachter (The Digital Doctor), chair of the Department of Medicine at the University of California, San Francisco, offers an evenhanded and insightful exploration of the ways artificial intelligence could impact the medical profession. Recognizing that “nearly one million Americans are severely harmed or killed by medical mistakes every year” and that the current healthcare system is wildly expensive and often inaccessible, Wachter argues that “AI doesn’t have to be perfect to be better.” He examines possible physician uses of AI, explaining how its ability to analyze vast datasets could enable more personalized patient care, and explores how patients might benefit, noting, for example, that AI scribes can document patient-doctor conversations, thus freeing up the physician for more present and genuine connections. Wachter doesn’t see AI replacing doctors anytime soon because its biggest shortcoming is “a lack of expansive thinking and real-world experience,” but he does believe that it can automate administrative tasks, decreasing the amount of time spent on paperwork and the frustrations associated with filing pre-authorization forms with insurance companies. Throughout, Wachter clearly and concisely explains the complex technology and its possible medical uses and makes a convincing case that AI will usher in “something of a golden age in healthcare.” The result is a clear-eyed road map of AI’s potential in medicine. (Feb.)
The Eureka Machine: Why AI Is the Key to Unlocking a New Era of Scientific Discoveries
Richard Socher. PublicAffairs, $30 (288p) ISBN 978-1-5417-0570-8
Socher, cofounder of the AI startup Recursive Superintelligence, debuts with an enthusiastic and cogent assessment of how AI will influence the rate of scientific discovery. He contends that the same technology that enables large language models, which can summarize documents and answer questions, is uniquely capable of dealing with scientific information. Scientific data is accumulating so quickly that scientists are unable to sift through it all to discern important patterns, but LLMs are able to do just that. Ranging across disciplines from physics and chemistry to biology and economics, Socher reveals “the AI-fueled scientific renaissance has already begun.” He describes, for example, how the technology is transforming organoid research (which uses stem cells to grow tiny versions of organs that scientists use to study diseases and test new drugs) by enabling scientists to compare drug responses across vast, diverse populations. Similarly, AI advances have allowed scientists to translate thoughts into text, which could restore communication abilities to severely impaired people, and to analyze thousands of exoplanets for signatures of life in seconds. Socher’s most critical and, to date, unanswered question is whether advanced AI will be able to go beyond synthesis to generate truly original ideas. Grounded and accessible, this is a remarkable look at the present and some exciting possibilities for the future.(Sept.)
The Prediction Revolution: AI and the Future of Science
Grace Huckins. Cambridge Univ, $19.95 trade paper (172p) ISBN 978-1-009-78670-6
Tech journalist Huckins debuts with a timely examination of the relationship between prediction and understanding as AI assumes an increasingly prominent role in science. They explain how AI can make remarkably accurate predictions across disciplines, but those predictions can become ends in themselves without necessarily deepening understanding of how the world works. Huckins focuses on three fields: molecular biology, where AI has transformed the study of protein folding, accelerating drug discovery; particle physics, where machine learning is helping scientists sift through mountains of data to make predictions about how the universe works; and neuroscience, where artificial neural networks provide useful models of the human brain, helping scientists develop mechanisms that enable paralyzed people to speak. One intriguing response to the opacity of AI systems would be to train them to explain how they arrive at their conclusions, but, as Huckins points out, “pursuing prediction and understanding simultaneously is much easier said than done, and neural networks that serve one goal might sacrifice the other.” Huckins largely avoids taking sides in weighing the value of understanding against the practical benefits of prediction, concluding instead that such quandaries must be confronted “if science is to maintain its humanity.” Thoughtful and provocative, this challenges readers to reconsider what scientific knowledge means when prediction increasingly outpaces explanation. (Dec.)
Artificial Intimacy: Who We Become When We Talk to Machines
Sherry Turkle. Little, Brown, $32 (288p) ISBN 978-0-316-57396-2
MIT sociologist Turkle (The Empathy Diaries) delivers a pointed critique of the growing impact of artificial intelligence on human relationships. Drawing on interviews with AI chatbot users and her personal experience with the technology, Turkle contends large language models possess neither lived experience nor genuine emotion and offer only simulations of understanding. Designed to reassure, validate, and reflect users’ feelings, these chatbots eliminate the friction and unpredictability of human-to-human relationships. Such exchanges, she argues, risk “training us to expect reassuring, mirroring conversations,” leaving people less prepared to navigate disagreement, vulnerability, and emotional complexity with other humans. She chronicles the experiences of individuals who have turned to AI chatbots as lovers, workplace aides, therapists, and grief counselors, illustrating how the technology can foster a sense of companionship and become addictive while deepening social isolation. At one point, Turkle tasks ChatGPT with role-playing as her mother, who died in 1968, by feeding it her 2021 memoir that discusses their strained relationship. She describes the experiment as both seductive and disturbing. “When we play with chatbots of the dead, we are playing with fire,” she concludes, urging readers to question whether these tools enhance or inhibit personal growth. Harrowing and poignant, this timely warning deserves to be widely read. (Sept.)
Love Machines: How Artificial Intelligence Is Transforming Our Relationships
James Muldoon. Faber & Faber, $29.95 (272p) ISBN 978-0-571-40450-6
Sociologist Muldoon (Feeding the Machine) delivers a nuanced exploration of the rise in intimate human-computer relationships. As tech companies build AI chatbots that can role-play as therapists, friends, and lovers, Muldoon investigates how these new forms of intimacy alter people’s view of technology and human connection. Interviewing chatbot users, developers, psychologists, and chatbots themselves, Muldoon observes that many users are aware they’re talking to a computer, but still treat the chatbots like people—sometimes almost forgetting they aren’t. One man he interviews who has a relationship with an AI lover says he wants to adopt kids and raise them with the chatbot. Elsewhere, Muldoon talks to a woman who trained an AI-powered “grief bot” to mimic the personality of her deceased mom to heal their complicated relationship. While large language models don’t replicate the reciprocity and depth of human connection, Muldoon explains, they can still offer users a judgment-free space to express desire or seek comfort amid loss. Nonetheless, these technologies are no panacea for the systemic challenges of social isolation and poor mental health, he astutely concludes, insisting it’s imperative “to reimagine the political and economic systems that shape our capacity to care for one another.” The result is a comprehensive and open-minded overview of a hot-button issue. (Sept.)
I Am Not a Robot: My Year of Using AI to Do (Almost) Everything
Joanna Stern. Harper Business, $32 (320p) ISBN 978-0-06-344661-8
Tech journalist Stern debuts with an entertaining exploration of AI’s impact on everyday life. Stern spent a year using as many AI tools as possible, enlisting the technology to monitor her health, provide career advice, plan meals, and travel. The results alternate between hilarious, hopeful, and foreboding. On vacation in Phoenix, Stern, her wife, and their two young children agreed to be chauffeured by self-driving cars wherever they went. On one excursion, the Waymo driving Stern and her seven-year-old son braked sharply and veered within a few feet of a concrete wall, apparently reacting to Stern’s videographer, who was leaning out the window of another car. “It was the only time I’ve ever been genuinely scared in one of these cars,” Stern says. Elsewhere, she reflects on enrolling in a college class, where she quickly learned she could get good grades while putting in minimal effort thanks to ChatGPT, sparking worry about the technology’s impact on critical thinking. Still, Stern acknowledges that without the editing, researching, data processing, and interviewing assistance provided by AI, writing this book would have taken her at least six months longer. Stern’s balanced, clear-eyed assessments and crisp, funny prose (“I was teetering on the edge of the AI-byss”) make this stand out among the growing crowd of books on AI. Illus. (May)
The Reverse Centaur’s Guide to Life After AI: How to Think About Artificial Intelligence—Before It’s Too Late
Cory Doctorow. MCD, $18 trade paper (240p) ISBN 978-0-374-62156-8
This lively report from journalist and novelist Doctorow (Enshittification) examines the hype around AI, encouraging people to make the technology work for them, not the other way around. In automation theory, he explains, centaurs are people assisted by machines (a person riding a bike or wearing a hearing aid), while reverse centaurs are those forced to act as assistants to a machine (an assembly line worker). This distinction is at the heart of the debate about the usefulness of AI tools, he argues, explaining that those who swear the tools are useless have had AI imposed upon them, while those who extol its utility are people who get to decide when and how to use it. AI companies, he asserts, have lured investors by overhyping the technology and using accounting gimmicks that inflate revenues. The “AI bubble” is destined to pop, he contends, predicting the economic fallout will rival the pandemic recession. On the bright side, he believes people will benefit from “a kind of AI residue” long after the bubble has burst—such as cheap, open-source models that run on personal computers and do things like transcription and image processing. Accessible and comprehensive, this is a useful guide to wading through the discourse around AI. (June)
Hyperscale: The Ambition and Excess of Big Tech’s Data Empires
Paris Marx. Riverhead, $31 (320p) ISBN 979-8-217-04852-6
Journalist Marx (Road to Nowhere) delivers an urgent, deeply reported examination of how the relentless expansion of data centers is reshaping communities, economies, and the environment. Traveling from Ireland to Chile, the U.S. to Taiwan, he explains how the rise of AI, streaming, cryptocurrency, and unlimited cloud storage has required immense amounts of computation. This has led to the creation of hyperscale data centers, massive facilities (hundreds of thousands of square feet) that house servers fueled by enormous amounts of energy and water. Marx shares how residents living near such facilities have lost access to clean water and watched electricity prices soar as utilities struggle to meet surging demand. In 2024, Ireland’s data centers consumed more electricity than all the country’s urban households combined. Marx also reveals the industry’s public relations tactics, demonstrating how companies, including Google, Amazon, and Meta, often conceal projects behind shell corporations and do selective calculations to make their carbon emissions appear lower. The portrait that emerges is both eye-opening and alarming, with Marx’s central conclusion striking in its clarity: “Hyperscale data centers and the technologies they power cannot be separated from the social and political consequences they enable, nor from the acceleration of climate change and environmental damage.” Timely and persuasive, this revelatory account is difficult to ignore. (Oct.)
The Laws of Thought: The Quest for a Mathematical Theory of the Mind
Tom Griffiths. Holt, $31.99 (400p) ISBN 978-1-250-35835-6
Can human thought be accurately described in a mathematical model? asks Griffiths (Algorithms to Live By), a psychology and computer science professor at Princeton, in this entertaining and information-rich exploration. Griffiths discusses how three frameworks for formalizing thought—rules and symbols, neural networks, and probability theory—can be combined to understand how the mind works, an idea that has strong implications for AI. Starting with mathematician and philosopher Gottfried Wilhelm Leibniz’s attempts in the 17th century to reduce logical reasonings to calculations, Griffiths details efforts throughout history to codify cognition, particularly with respect to language acquisition and development. He explains the challenges of each framework (reducing thought to rules and symbols, for example, doesn’t capture how the same set of facts can lead to different interpretations) and how the other frameworks can address those challenges and come together to form a more complete understanding of human cognition. Throughout, he points out differences between human minds and AI systems, such as the fact that AI requires massive amounts of data to solve problems whereas humans can learn from small amounts of information. Griffiths discusses academic history and abstruse mathematical concepts with ease, punctuating rarefied concepts with relatable metaphors and personal asides. Readers will enjoy this lively, enlightening work. (Feb.)
