Maybe the World Will Muddle Through Somehow - Los Angeles Review of Books

The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence by Sebastian Mallaby. Penguin Press, 2026. 480 pages.

SO MUCH OF THE hand-wringing and existential dread surrounding artificial intelligence and how it has enveloped our lives has to do with not just the technology itself—increasingly powerful, malleable, and ubiquitous—but also the people in charge of it. The outsize influence of moguls like Sam Altman, Elon Musk, and Mark Zuckerberg has shocked even a generation that grew up watching oligarchs usurp vast swaths of industry. It is thus a surprise to see Demis Hassabis, another billionaire, primarily portrayed as a man not only disinterested in but also actively eschewing the accumulation of wealth and power. “The worst thing you can do to somebody is to be controlling […] I go to great lengths not to be like that,” he tells his biographer Sebastian Mallaby, whose new book, The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence, lays out how much control over our lives Hassabis may inadvertently have.

Hassabis’s home office is an attic without a view; he has no ski chalets, beach houses, or yachts; and his few indulgences include attending his favorite club Liverpool’s soccer games for £3,000 a year. A “lovely” and “phenomenally bright” kid, according to an early colleague, he became a chess master at 13, using that analytical skill to design machines that could beat gaming champions. He relentlessly chased artificial intelligence, aware of its power to unlock a prosperous future or destroy it, driven always by the sheer joy of discovery. His research won him the Nobel Prize in Chemistry in 2024.

The CEO of Google DeepMind is one of the earliest names behind the AI boom, and behind the theory that machines can and will mimic the human brain in connections and capacity (and blunders). Hassabis is a CEO by title and function but a reluctant one, his business activities conjoined with the raw curiosity of the scientist. He also has an appetite for pain: at the age of 10, he defined doing his best as pushing himself “to the point just before death. Because that is literally when you have done your best. If you die—by die, I mean burn out or something—then you’ve slightly overdone it.”

Along with Altman and his partner-turned-rival Dario Amodei (who runs Anthropic, the $380-billion lab behind the chatbot Claude), Hassabis can be seen as the third figure in today’s AI triumvirate, although the book tells us that perhaps he was the first. Given Altman’s and Amodei’s prevalence in the headlines—for, among other things, haggling over US defense contracts, displaying sociopathic tendencies, indulging in misleading business practices, and showing such a competitive spirit that they avoid physical contact in public forums—Hassabis might be the least understood of the trio.

Mallaby, whose last book was a best-selling history of venture capital, is skilled at tracing the motivations and folkways of the powerful. His 2016 biography of Alan Greenspan, the Federal Reserve chairman who was blamed for the 2008 financial crisis, is titled The Man Who Knew, a sign of its author’s keen eye for drama and irony. Compared to typical journalists, whose nonfiction books stem from years of beat reporting and frequently topsy-turvy (if not adversarial) relationships with their subjects, Mallaby employs a different approach: he convinces key subjects to give dozens of hours of interviews (Hassabis gave him 30), tracking them intently over many years. Typically, this risks surrendering control of the narrative to the subject, who can extend or withhold access, but Mallaby has deftly navigated this dilemma, and his books as a result display an admirable balance.

Despite Hassabis’s genius, or perhaps because of it, he was late to recognize the role of chatbots in spurring AI’s penetration into our lives. Reading this book alongside the current technology news cycle can feel like encountering a key but somehow forgotten historical figure—a misleading impression given how ably Google has wrestled its way back to relevance.

Born in 1976 in North London to a Chinese Singaporean mother and a Greek Cypriot father, Hassabis as a toddler and teen swept chess games in a fiercely competitive circuit where parents berate their children if they lose (as Hassabis’s father did in one instance). There was a “thin line between exultation and breakdown,” Mallaby writes. For the longest time, the boy assumed that chess would be his career, the cerebral game driving his desire to unpack the human brain.

Peter Molyneux, an intense and magnetic game developer, hired the 16-year-old Hassabis, who had to bide his time before enrolling at Cambridge University. Molyneux gave Hassabis a copy of Gödel, Escher, Bach: An Eternal Golden Braid (1979), the Pulitzer-winning tome about computers, language, AI, and free will, which changed the course of his life. In the early 1990s, for a video game called Theme Park, Hassabis crafted lively details: if players put extra salt on their fries, cold-drink sales would explode, while roller coasters that were too scary would induce vomiting.

Hassabis went from designing games to designing technology to beat top gamers, as part of a growing cohort of scientists who felt that the best way to test artificial intelligence was to combine rules and creativity in just the right mix. Based in London, and now armed with a PhD in neuroscience, he founded DeepMind in 2010, a start-up dedicated to figuring out artificial general intelligence (AGI)—the hypothesized phenomenon of machine intelligence surpassing humans in all fields. He was convinced this achievement would drive social progress in a world that had no immediate answers to the existential questions thrown by the financial crisis and the swelling global population.

DeepMind’s researchers believed that the best way toward artificial intelligence was for machines to connect dots in the style of the human brain, using so-called neural networks that mimic our brain’s pathways, as well as reinforcement learning, a practice whereby a machine learns through intense, repetitive feedback loops and reward signals, rather than simply being fed troves of data. Hassabis had the charm and determination that marks out influential people but also an apparent purity of purpose that tends to dwindle among the powerful as commerce comes to dominate. The founder of every AI start-up claims this sort of purity, but Mallaby provides more evidence for Hassabis than anyone can conjure for, say, Altman or Zuckerberg.

Despite an aversion to corporate hierarchy and Silicon Valley insularity, Hassabis sold DeepMind to Google in 2014 for $650 million because he saw this as his best shot at developing AI at his pace, using their resources. At the time, Google’s venerable motto, “Don’t Be Evil,” was hanging in the balance but didn’t yet ring hollow.

This sets up the irony and tension of the book, and of Hassabis’s life and technology writ large. Can the most brilliant, well-intentioned people drive societal change from a position of corporate power without their shifting incentives swallowing them whole? Even before the Google deal, Hassabis and his cofounders and friends were aware of the consequences if AI development went wrong. They still don’t have solutions to the problems they are birthing but believe that their mere awareness will serve them well. As long as they are the masters of their choices, and as long as AI development doesn’t just become a competitive race, they can live in peace.

Hassabis’s belief differed from that of billionaire investor and LinkedIn cofounder Reid Hoffman, who had first bet on an upstart called OpenAI in 2015 but later committed a billion-dollar fortune to DeepMind. Hoffman felt that multiple AI labs could separately gun for success, similar to a multiparty democracy in which, Mallaby writes, “this pluralism would be balanced by a shared commitment to bedrock values.” In the years since, Hoffman’s hunch has largely prevailed—but the “bedrock values” have frayed, not just in tech but also in the very democratic systems that inspired the analogy. That tension—between optimism and unease—finds a neat distillation in the words of transhuman theorist Ray Kurzweil, who captures the enduring reaction to superhuman intelligence in three beats: “Wow!, Uh-Oh, and What Other Choice Do We Have but to Move Forward?”

A series of big breakthroughs came between 2015 and 2017, when DeepMind mastered the 4,000-year-old board game Go, which is far more complicated than chess. DeepMind’s program AlphaGo beat successive champions, including one in Seoul in a match Google’s former CEO Eric Schmidt and cofounder Sergey Brin flew in to watch. The match drew 200 million viewers—more than the Super Bowl and more than double the audience that, two decades prior, had seen chess titan Garry Kasparov lose to a machine, a seismic moment in tech history.

These successes stemmed from DeepMind’s focus on making technology similar in behavior to the human brain, and they came after and amid years of experiments and gaffes—for example, a machine playing Pong that merely “bobbed about indifferently” in different parts of the screen. That leap of faith and imagination foreshadows the biggest problem with AI today: hallucination. When ChatGPT presents as fact the names of books that don’t exist and quotes that were never said, it is magnifying a foundational flaw. The human brain can often be tricked into “reconstructing a memory of something that had never happened,” the so-called Mandela Effect.

In the process of unpacking each piece of AI development, Mallaby ruminates on how our brains work. Specifically, our “intelligence” lies in our ability to link unrelated facts and channel them into a program for daily existence. If AI is truly to surpass me, it has to know everything about what I see, eat, drink, watch, think, and do every day. These descriptions of how scientists strive to make machines imitate the brain serve as paeans to human intelligence themselves, while reinforcing the slow, incremental nature of progress that eventually changes the world.

In 2017, OpenAI’s systems, after digesting reams of Amazon reviews, started generating a “sentiment neuron,” which fired or remained dormant depending on whether the review was positive or negative. It isolated the concept of sentiment to a single technological piece, without even being told to do so. Developments like these show the disparate yet momentous nature of the AI boom we see today, raising the question of what else may be happening in tech labs today that will upend our lives in the coming months and years?

Mallaby’s narrative lays out the warring, messy philosophies that determine AI’s path. Is building AI tantamount to complicity in all of the technology’s downsides? Or is watching from the sidelines as someone with contrasting values builds it worse? Is safety a governance challenge, to be solved by regulation and corporate guardrails, or a technical challenge, to be encoded into the machines as nonnegotiable limits? The book, smartly, doesn’t offer solutions to these fraught questions so much as it lays out the challenges and tradeoffs, in the process mirroring the decision-making of the characters whose lives it chronicles.

Hassabis pursued AI as a way to understand the world and solve complex problems in physics, chemistry, and biology. He avoided large language models (the tech that powers ChatGPT, Claude, and Google Gemini) because he believed that simply processing all the text on the internet would yield underwhelming results. But his instinct, even if right, might not matter because LLMs are now the entry point into more sophisticated uses of AI. In the decade after selling to Google, Hassabis navigated considerable corporate politicking: planning to spin off DeepMind and then instead “spinning it in” to a more central role within Google; balancing his vision for AGI and scientific discovery against Google CEO Sundar Pichai’s acute awareness that AI would disrupt Google’s search business, while keeping DeepMind relevant to that purpose; watching OpenAI, founded as a laggard, leapfrog DeepMind and become the de facto AI company; and contending with Elon Musk’s hunger to seed OpenAI explicitly as a counterweight to Google—his prescient but erratic and often toxic interventions making him a force impossible to ignore or trust.

But in 2024, Hassabis co-won the Nobel Prize in Chemistry for his AlphaFold program, which used AI to make breakthroughs in protein folding—predicting the complex shapes that proteins assume inside our bodies, a discovery that can potentially unlock cures for Parkinson’s and Alzheimer’s. A year later, Mallaby writes, “more than three million investigators across the world had freely consulted AlphaFold’s predictions, accelerating their work in everything from fundamental biology to vaccine development to environmental sciences.” Efforts that once took years now took minutes. Later on, Hassabis reflected that “science is where AI does unequivocal good, […] whereas with language models there can obviously be bad use cases.”

The launch of ChatGPT in 2022 forced Hassabis’s hand, and Google jumped headfirst into the AI rat race, launching and refining Gemini. The dogmatism that led him to the Nobel Prize also put his status as an AI pioneer under serious threat. Google’s language models eventually even beat OpenAI in tech prowess, becoming both a comeback story for the ages and a lesson on how tech companies with hundreds of billions to spend can seem like the cat with nine lives. In the last 12 months, Google’s market value has doubled to four trillion dollars.

J. Robert Oppenheimer, the physicist who built the atomic bomb and spent his later years ruing its ripple effects and devastation, is mentioned at least 12 times in The Infinity Machine. The book nods to the fact that Sam Altman shares a birthday with Oppenheimer, but it is Hassabis who more fully inhabits the moral tension Oppenheimer epitomizes, bound together as they are by origins in science rather than commerce, and by their willingness to confront the social uses of their own work.

Mallaby captures the lure of discovery—“the Icarus instinct”—that overwhelms every innovator’s better judgment, and Hassabis himself acknowledges the paradox:

It should feel amazing, realizing all these dreams […] But it doesn’t feel like how I imagined it would feel. The way it’s going, right, this mad rush.

So I’ve had to make my peace with that. Recognize that it’s going to be messy, and I’ll just have to do the best I can. And maybe we, being the world, will muddle through somehow. I’m optimistic still.

Ultimately, this is a tale of how capitalism warps incentives even for those who enter it with clear eyes. Hassabis chose Google, calculating correctly that proximity to resources and scale was the only way to pursue AGI seriously. That this path led him into an arms race he never wanted, and to mixed feelings about what he has helped unleash, is the central irony of his story. And yet: the Nobel Prize, the science, the breakthroughs—it is hard to argue that he was wrong.

Weighed against a gallery of larger-than-life technologists who have prompted literary explorations—Musk, Altman, Sam Bankman-Fried—Hassabis emerges as a rare and oddly likable figure: someone who, like most of us, has had to operate within structures that reward things he does not fully believe in, and who has done so without losing himself entirely.