This article is taken from the June 2026 issue of The Critic. To get the full magazine why not subscribe? Find our subscription offers here.
Britain wants its boffins to be as humble as Alec Guinness in an Ealing Comedy, and Sir Demis Hassabis obligingly fits the bill. The ferociously clever and competitive entrepreneur and computer science researcher, awarded the Nobel Prize in 2024 at the age of 48, has cooperated at length with Sebastian Mallaby, the author of this book.
The title hints that this is not a straightforward biographical work, however. The “quest” is the mission to create a God-like superintelligence, and it is this that gets star billing. “Artificial general intelligence”, or AGI, is a hypothetical computer that is infinitely wise, solves riddles and makes discoveries that we cannot.
Mallaby assures us that this has been Sir Demis’s life-long goal. Hassabis speculates that once invented, it would remove the need for money.
Hassabis is not the only Captain Ahab in search of this white whale. Today’s equity bubble, accompanied by the largest splurge of capital in history, is founded on two assumptions, and the proximity of AGI is one of them.
The second assumption is rather more cynical: which is that even if AGI is never achieved, there will be sufficient advances in artificial intelligence to collapse the value of a number of other economic sectors, with the beneficiaries being a small handful of technology companies. Google will tell Tesco how to price bananas, for example, or replace an osteopath’s customised exercise plan. A small number of technology companies will capture this value instead, becoming dominant supercompanies.
Critics may ask: even if AGI was feasible, would it be desirable? It is difficult to envisage a major economy trusting an AI to make major decisions, such as setting interest rates, or to imagine governments expecting it to write peace treaties that signatories would respect.
However, few serious practitioners take the prospect of imminent AGI seriously. The doubters include leading roboticists and computer scientists, ranging from Turing Award-winner and AI pioneer Judea Pearl to deep learning guru Yann LeCun. Computer superintelligence was a fringe concern until money from Peter Thiel, Elon Musk, Google founder Larry Page and the rationalist movement found it, and took it mainstream. Hassabis was here early, leveraging this elite social network to fund his DeepMind startup, displaying a persuasiveness and talent for showmanship that has recurred throughout his career.
Unfortunately, whilst Mallaby is a skilful storyteller, dividing the reader’s attention between Hassabis’s career and AGI compromises his enquiry. If AGI is a cynical exercise in deception, what is a character as compelling as Hassabis doing selling it? Does he really believe in it? Alas, such qualms are not permitted to sully the hero’s journey.
We can at least be grateful to Mallaby for unearthing more biographical detail about an extraordinary and restless intellect. Born in 1976 in North London to a bohemian Greek Cypriot father and a Singaporean shop assistant mother, Hassabis was playing chess competitively from the age of six, becoming a master at 11.
Accepted into Cambridge two years early, he instead took a gap year at the flamboyant computer games studio Lionhead; he would later found his own games company. Burned out by business and the lavish and unfilled promises he had made as CEO, he turned to academia, publishing an attention-grabbing neuroscience PhD thesis in 2009, which proposes a new theory of memory construction. His mysterious AI research lab, DeepMind, was acquired by Google in 2014 for £400 million. It made no products: Google was buying the man.
DeepMind produced a number of eye-catching showcase demonstrations: it learned 1980s computer games, then beat a human Go player, and accelerated the discovery of promising potential protein structures, which earned Hassabis global acclaim. Readers may be surprised to find that Hassabis’s theory on memory formation has not held up well, and that the number of proteins discovered by AlphaFold is so far zero.
Mallaby recounts several incidents demonstrating that Hassabis is compellingly persuasive, capable of dazzling investors and potential PhD hires. His co-founder Mustafa Suleyman offers a rare note of scepticism: he is dismayed by the ease with which Hassabis can dissemble in front of his employees.
If he were not pursuing AGI, one wonders what useful work someone as brilliant as Hassabis might be doing otherwise. What lasting, tangible marvels could he apply himself to creating instead?
