Plenty of books, articles, and essays about so-called artificial intelligence are a bit boring. They may be accurate, knowledgeable, worthy, even wise, but they often lack the key elements of drama that make writing captivating: conflict and resolution, expectation and surprise, the general and the particular, comedy and tragedy. They’re long on exposition, analysis, and prognostication but short on narrative, as if written for the classroom rather than the beach.
Happily, the opposite is true of two recent blockbusters on the subject: Karen Hao’s Empire of AI: Inside the Reckless Race for Total Domination and Sebastian Mallaby’s The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence. In many ways, the books are very different, as you can tell from their subtitles.
Hao tells the story of Sam Altman and OpenAI, which is a fundamentally pessimistic story about the triumph of money over idealism, full of pantomime villains and adolescent Silicon Valley squabbles and the devastating effects of Western techno-imperialism on Majority World communities. Idealism has a much better time of it in Mallaby’s book, which follows the chess prodigy and Nobel Prize–winning scientist Demis Hassabis—described by a leading British journalist recently as “greatest Briton alive”—from North London to Cambridge to Google DeepMind.
The individuals, companies, models, ambitions, technologies, and motivations in the two books are strikingly different. So are the answers to two crucial questions: Is artificial general intelligence (AGI) possible? And on balance, is that a good or a bad thing?
But both books have the same secret sauce. They’re marvelously readable, and often gripping, for the simple reason that they tell a fantastic story. There is, of course, lots of detailed exposition (How exactly does this technology work? Who is trying to get it to do what?). There is rigorous analysis (How is it developing? What are its consequences?) and thoughtful prognostication (Where might it go next? What might that mean for humanity?). But at root, both books are telling a good old-fashioned story—a tale of individual human beings, with all their gifts, flaws, hopes, fears, successes, failures, and longings—and it’s the story that holds the argument, and the reader’s attention.
Looming Colonialism
Empire of AI begins with the 2023 board meeting in which Sam Altman, CEO of OpenAI and driving force behind the launch of ChatGPT, is told he’s getting fired. The firing barely lasts the weekend. Despite losing the confidence of the board that oversees the company, for a pattern of misleading and “not consistently candid” communication, Altman is quickly reinstated after a groundswell of complaint, an intervention from Microsoft, and a threat of mass resignations from the staff.
In Hao’s telling, the botched ouster serves as a parable of the tensions within OpenAI, and indeed Altman himself: between altruistic humanitarian ideals and grubby commercial realities, independent nonprofit oversight and moneyed corporate interests, the “Open” of the company’s original vision and the billions of dollars and expanses of land needed to build “AI.” Many of these tensions exist, of course, within the AI industry as a whole.
The narrative that follows goes down three tracks. One is technological, describing how large language models (LLMs) emerged to become the dominant expression of AI in the 2020s. Even in 2020, it was by no means obvious that this particular form of technology would prove as popular and simulate human intelligence as convincingly as it has in the last five years, especially since the game-changing launch of ChatGPT.
Hao explains how this happened with admirable clarity, avoiding the dangers of meaningless hand-waving on the one hand and hypertechnical alphabet soup on the other. It’s refreshing to me, as an arts graduate and relative technophobe, to read an account of a fiendishly complex technology being developed through many iterations and still be able to understand it.
Even in 2020, it was by no means obvious that AI would prove as popular and simulate human intelligence as convincingly as it has in the last five years.
The second track is commercial and interpersonal. It centers on a number of brilliant, infuriating, egotistical, superwealthy, visionary, and/or paranoid individuals and the rivalries between them. This is where the subtitle (Inside the Reckless Race for Total Domination) comes in. Hao doesn’t doubt that the battle between altruistic hopes and financial power is so far being won by the latter, at least in the case of OpenAI, and she makes a compelling case. But she’s a good enough journalist and storyteller to help us see things from the perspective of each actor, even if some (most prominently Altman and Elon Musk) are clearly the villains of the book.
But the third track—and this provides the book’s most distinctive contribution, along with its title—is what we might call imperial. It moves around the world from Bogotá to Nairobi to Santiago, revealing the (sometimes chilling) human cost of AI. In Hao’s words,
The empires of AI are not engaged in the same overt violence and brutality that marked [colonial] history. But they, too, seize and extract precious resources to feed their vision of artificial intelligence: the work of artists and writers; the data of countless individuals posting about their experiences and observations online; the land, energy, and water required to house and run massive data centers and supercomputers. . . .
OpenAI is now leading our acceleration toward this modern-day colonial world order.
Hao’s predictions are ominous but well supported.
Hopeful Horizons
Readers for whom that all sounds a bit apocalyptic might prefer an AI palate cleanser: a portrait of a more noble individual, a more idealistic company, and a more beneficial sort of technology. That’s what Sebastian Mallaby has provided in The Infinity Machine. (While I was writing this review in a coffee shop, the young woman sitting next to me asked me about the two books; I described them both, and then saw her take a photo of Mallaby’s but not Hao’s. “The technology is inevitable, so I think I’ll read this one,” she said.)
Hassabis is undoubtedly a much more sympathetic figure than Altman, and I don’t think I’m only saying that for patriotic reasons. He’s an extraordinary individual. We encounter him as a chess prodigy, game designer, theoretical physicist, neuroscience researcher, computer scientist, Mind Sports world champion, and Nobel laureate rather than as a tycoon aiming for world domination. In Mallaby’s telling, Hassabis became an entrepreneur because it was the best way of achieving his scientific and humanitarian objectives, and still cares far more about the science than about the wealth or power it produces.
Consequently, The Infinity Machine has more to say about the possibilities of designing games and folding proteins than of raising capital and poaching talent. This gives the book a much more optimistic flavor:
The good news is that Demis Hassabis, who blazed the trail followed by rivals, is decent and public-spirited and wants the best for humanity. He has ego, to be sure. He is fearsomely competitive; his sense of destiny, as the developer of AI, borders on the messianic. But his goal is scientific enlightenment, not money or power. . . . The overweening self-righteousness of the most voluble tycoons has discredited the idea that messianic innovators can possibly be well-intentioned. But there is no necessary connection between making a fortune in the Valley and egomania unbound, and nor is it fair to paint the technology sector as a bastion of anarcho-libertarianism. . . . In these volatile circumstances, Hassabis’s struggle to stay true to his personal values is a defining story of our times.
Here’s another way of putting it. In places, Empire of AI is a bit like reading Robert Caro on Lyndon Johnson, in which the overriding question is roughly this: “How on earth did this unscrupulous person end up with so much power?” The human story is powered by intrigue and manipulation; the reader hopes against hope that the protagonists will not succeed, even though we know they will.
By contrast, The Infinity Machine is more like reading Walter Isaacson on Benjamin Franklin, where the reader is asking, “How on earth did this brilliant person come up with so many good ideas?” The fascination is theoretical and intellectual, not political or economic. We’re rooting for Hassabis’s approach to AI—based more on deep learning and mastering complex games than on LLMs and word prediction—to beat OpenAI to the punch, even though we know that, at least for the first couple of rounds, it won’t.
Alternative Futures
This more positive perspective shapes Mallaby’s narrative of AI as a whole. His book largely avoids the moral and ethical downsides to which Hao gives so much attention, like the environmental costs of data centers or the human costs of labeling explicit content, and talks much more about the upsides, from medical breakthroughs to eliminating microplastics.
Some of this simply reflects the different personalities of Hassabis and Altman; if you judge that AI is being pioneered by idealistic scientists, you’ll be less jaundiced about it than if you think it’s being led by scheming billionaires. Some of it reflects the different approaches and achievements of DeepMind and OpenAI. After all, unlike Hassabis and his AlphaFold program, ChatGPT hasn’t gotten close to doing anything so useful as folding a protein and winning a Nobel Prize.
If you judge that AI is being pioneered by idealistic scientists, you’ll be less jaundiced about it than if you think it’s being led by scheming billionaires.
Mallaby is also much more open than Hao to the idea that superintelligence is actually possible, not least because Hassabis himself is so committed to the concept. Arguably the most exciting section of his book is the chapter on how Hassabis’s AlphaGo program first defeated the best human player at the ancient strategy game Go, which has 324 squares and makes chess look as simple as tiddlywinks. In the course of the contest, AlphaGo made a totally counterintuitive and unpredictable move, now known simply as Move 37, which baffled observers and took its human opponent a full 12 minutes to respond to, but turned out to be astonishingly far-sighted and strategically decisive.
Mallaby freely uses words like “intuition” and even “superintelligence” to describe what happened. Hao is far more cautious, in part because there’s no agreement on exactly what intelligence even is. “The name artificial intelligence was thus a marketing tool from the very beginning,” she explains.
So who is right? Is AGI a real possibility, the most transformative innovation of the last 70,000 years, and a breakthrough with the potential to enhance all our lives by solving problems humans can’t, under the guidance of quirky, obsessive, and brilliant scientists? Or is it a chimera, a tech-bro vanity project, a nicely branded way of describing predictive text on a massive scale, that allows rich people to work (and perhaps cheat) more efficiently while poorer communities pick up the tab?
Claude would presumably tell me what an interesting question this is, and how elements of both are involved, depending on all sorts of variables. I’m less equivocal. Temperamentally I’m a optimistic person, but I’m substantially more convinced by Hao’s read of the way the technology is going—to say nothing of the financial, technological, and even geopolitical incentives—than Mallaby’s.
I hope I’m wrong. I hope the scientific discoveries that will undoubtedly follow will compensate for, or even obviate, many of the downsides. But as of 2026, we look closer to an Empire of AI than we do to an Infinity Machine.
The Gospel Coalition