There is a scene early in Sebastian Mallaby’s new book The Infinity Machine: Demis Hassabis, DeepMind and the Quest for Superintelligence that may unsettle anyone who has taken comfort in the idea that the people building AI (artificial intelligence) know what they are doing. The Infinity Machine: Demis Hassabis, DeepMind and the Quest for Superintelligence, Sebastian Mallaby, Penguin Books, 2026.Geoffrey Hinton, Nobel laureate, celebrated as “godfather of deep learning,” and considered one of the most revered minds in modern science, is at the Royal Society in London. Spotted in conversation with the philosopher Nick Bostrom, Hinton confesses that he believes AI systems will be weaponised against ordinary people. When Bostrom asks why he is building them anyway, Hinton’s answer is disarming in its honesty: the prospect of discovery, he says, “is simply too sweet.”That exchange gives Mallaby his title and an animating question: Is the race to build artificial general intelligence a story of scientific heroism, or a tragedy dressed up in the language of progress? The book does not resolve this question, perhaps it cannot, but it is a formidable attempt to hold it open without flinching.The protagonist of the book is Demis Hassabis, the co-founder of DeepMind and winner of the Nobel Prize in Chemistry in 2024, whose company was acquired by Google and has now become the most scientifically credentialed AI lab anywhere. Hassabis, to say the least, is an exceptional individual. A child prodigy and neuroscientist, his doctoral studies of memory and imagination became the foundation for machine learning algorithms.Having already written some of the best books about hedge funds, the World Bank, and venture capital, thus putting himself among the great storytellers about the intricacies of capitalism, Mallaby spent three years and more than thirty hours talking directly to Hassabis to complete the book. The effort is obvious from the outset. The Infinity Machine flows with the pace of a thriller and the rigour of research. Mallaby spoke with more than a hundred people in and out of DeepMind, from the alienated co-founder Mustafa Suleyman to the architect of AlphaGo, David Silver.What makes Mallaby’s work quite special is how it weaves three separate stories into one narrative tapestry. First, there is the story of Hassabis himself. Mallaby is keenly aware of the paradoxical nature of his subject. “Power is not for me an interest in and of itself,” Hassabis once said to Mallaby, yet he managed to quietly oust Suleyman from the company they had founded together, using the cover of a misconduct investigation where a simple discussion could have sufficed.Also read: I Was Never Good at Exams. Was I a Bad Student?The second story follows the evolution of DeepMind, from its inception in a small London attic in 2010, when venture capitalists laughter off the ambition of constructing something approaching human-level intelligence, to the firm’s present status as the powerhouse behind Google’s ambitions in the field of AI. Mallaby does an excellent job chronicling the atmosphere inside the company, and its complex, often fraught integration into Google.The third narrative is the science, and here Mallaby does something genuinely useful. He explains, with clarity that rarely condescends, how reinforcement learning and deep learning were combined at DeepMind to produce systems that could learn by doing rather than by being told, systems that navigated an almost “infinite” space of possible moves to master “Go,” that folded “proteins” without being given the rules of chemistry, that began, in the words of the book’s title, to function as “infinity machines.” The account of AlphaFold, the system that predicted the three-dimensional structures of nearly all the proteins, a problem that had stumped structural biologists for fifty years, is particularly well rendered. The speed with which Hassabis turned from celebrating AlphaGo’s victory to declaring, on a Seoul street at midnight, that protein folding was now within reach is one of several moments where the book captures the peculiar restlessness of the very brilliant.What distinguishes The Infinity Machine from the crowded shelf of AI books is Mallaby’s refusal to let the safety question slide into the background. He returns to it repeatedly, and with increasing urgency as the book nears its end. The failure of every governance experiment, from the Independent Review Panel for DeepMind Health to OpenAI’s non-profit board, from the Bletchley Park summit to Hassabis’s own dream of a CERN-like international body to oversee the final steps to AGI, is documented with the weary precision of someone who has watched idealism meet institutional reality and come off second best. By 2025, as DeepSeek’s emergence signalled that the AI race had become genuinely multipolar and therefore essentially ungovernable, Hassabis speaks with uncharacteristic directness at Davos: “The agentic era we are about to enter into is a threshold moment for the systems becoming far more risky.” Hassabis himself concedes that the safety agenda he once championed through formal governance mechanisms has given way to something more personal: he will try to be in the room where decisions are made. It is the best comfort available. It is not much.Indian readers will find reason to attend to this book. The question of who governs transformative technologies, and whose values get encoded into systems that will increasingly determine how governments process information, how economies allocate resources, and how scientific research is directed, is not abstract. In a world where a handful of laboratories in London, San Francisco, and Shenzhen are racing to build systems of near-omniscient pattern recognition, the rest of the world is largely a spectator. Mallaby does not dwell on this asymmetry, but it is present in the book’s silences. The Global South does not appear as an actor in this story; it is, at best, a future beneficiary of protein databases and medical diagnostics, and at worst, an afterthought in a civilisational gamble being conducted without its consent.There are places where the book strains. Mallaby admires Hassabis, and the admiration sometimes tips into advocacy. The portrait of Altman is less nuanced than that of his subject; the treatment of Suleyman, whose own account of his departure from DeepMind is sharply at odds with what Mallaby describes, has the feel of a case already decided. Also read: Will AI Preside Over the End of Literature?The book also sidesteps certain uncomfortable questions with uncommon agility. The intellectual property rights of the writers, artists, and scientists whose work trained these systems are mentioned briefly; the labour practices that sustain the data centres and the human reviewers who shape model behaviour are largely absent. These are not peripheral concerns. They are part of the same moral architecture that Mallaby wants to examine.But these are all objections of a book that has earned its right to be taken seriously. The Infinity Machine succeeds at what great narrative nonfiction books succeed at: making us feel the gravity of events that, in real time, happen too quickly and are too technical for us to grasp. The phrase from Oppenheimer, “When you see something that is technically sweet, you go ahead and do it,” pops up in Mallaby’s work like a reproach. Hassabis knows the peril. He sat down at his desk at two in the morning and could feel, as he tells Mallaby, “like reality is staring at me, screaming at me…trying to tell me something if I could just listen hard enough.” He built the machine anyway.That combination of lucidity and compulsion, of knowing and proceeding regardless, is the defining condition of our moment. Mallaby has written about it with intelligence, honesty, and considerable grace. Whether the machine his subject has built will ultimately be kind to the species that made it is a question this book wisely declines to answer. The sweetness of the discovery has been had. The argument about what to do now has barely begun.Deepanshu Mohan is dean and professor of Economics at O.P. Jindal Global University. He is Visiting Research Fellow at the University of Oxford’s Department of International Development and Visiting Professor at London School of Economics (LSE).Aman Chain is a senior research assistant with Centre for New Economics Studies (CNES), O.P. Jindal Global University, and studies Law there.