What do you think?
Rate this book
無限機器：Gemini推手哈薩比斯的超級智慧長征
掌握Google大腦的諾貝爾獎得主
跳脫矽谷模式，打造超級智慧的無限機器
甫上市即榮登《紐約時報》暢銷榜、英國《週日泰晤士報》暢銷榜、以及中國《財新》暢銷書單。
★比人工智慧更「智慧」的終極發明
當科技巨頭大打AI戰時，哈薩比斯依然秉持「科學家」的使命，引領人類邁向前所未有的豐饒時代。
他手中掌握Google的AI技術命脈，心中牽掛如何打造出人類所能想像最強大的工具：通用人工智慧（AGI）。這會是一臺能在無窮無盡的資料之海縱橫馳騁的機器，應用範疇沒有極限，是解決所有問題的萬能解方。
★從西洋棋神童到AI界的賈伯斯
即便從天才輩出的科技業人才標準來看，哈薩比斯仍是異數。他五歲就成為西洋棋神童，十七歲共同創作出熱銷百萬的電玩遊戲，隨後進入劍橋大學電腦科學系就讀，跨界取得神經科學博士學位。為了揭開生物學與物理學的奧祕，他與志同道合的夥伴創立DeepMind，要打造一種能夠像人類一樣思考、推理、學習的智慧機器。
為了得到更多算力支持，他接受Google的收購以加速研究進展。自此，他擁有改變的影響力，領導Google的人工智慧研究，以AlphaGO打敗世界棋王，催生能與ChatGPT一爭高下的Gemini，更以AlphaFold研究摘下諾貝爾化學獎桂冠。
★採訪權限空前的卓越紀實
本書由兩度入圍普立茲獎的當代頂尖財經作家馬拉比，歷時三年、絕無僅有的三十小時促膝長談，深入剖析科技先驅德米斯．哈薩比斯，額外採訪超過一百位相關重要人物，詳實記錄DeepMind這家足以顛覆世界的公司發展。隨著AI對人類影響愈來愈深，這將成為定義我們這個時代最重要的篇章。
各界推薦
才華洋溢的哈薩比斯與他共同創辦的DeepMind，為正在改變世界的人工智慧開闢新的道路。這本報導詳實、洞察入微的著作中，記錄了技術突破、安全爭議、解開蛋白質摺疊之謎的輝煌勝利，以及與OpenAI等競爭者的激烈角力。這不只是一個精彩故事，更是AI發展歷程的深刻解析。
——華特．艾薩克森（Walter Isaacson），《賈伯斯傳》與《馬斯克傳》作者
本書捕捉了這場高風險AI軍備競賽令人屏息的戲劇張力。作者講述哈薩比斯如何建立DeepMind、引爆AI革命，並重繪全球權力地圖。這是在機器智慧的最前線，人類野心與倫理規範激烈碰撞的精采紀實。
——克里斯．米勒（Chris Miller），紐時暢銷書《晶片戰爭》作者
本書生動描繪哈薩比斯的多個面向，也是DeepMind最具權威的記述──這無疑是AI時代影響最為深遠的收購案。馬拉比用全新的視角來看塑造這個實驗室的人物性格、內部的明爭暗鬥，以及與Google的激烈角力，讓我們看到這個教機器思考的組織如何形塑而成。
——派翠克．麥奇（Patrick McGee），紐時暢銷書《蘋果在中國》作者
作者的高明之處在於，透過哈薩比斯這位當代最重要的技術變革者來講述人工智慧的壯闊篇章。本書不僅讓你深入了解AI，也讓你看見人類的才智、遠見與百折不撓的毅力如何翻轉這個世界。
——安德魯．麥克費（Andrew McAfee），《第二次機器時代》共同作者
作者以其一貫的非凡文采，講述一個迷人的故事：一個流著新加坡華裔與希臘裔塞普勒斯血液的北倫敦青年，如何引領Google在AI競賽領先群雄，還在途中摘下諾貝爾獎的桂冠。
——麥可．莫里茨（Michael Moritz），紅杉資本董事長
跳脫矽谷模式，打造超級智慧的無限機器
甫上市即榮登《紐約時報》暢銷榜、英國《週日泰晤士報》暢銷榜、以及中國《財新》暢銷書單。
★比人工智慧更「智慧」的終極發明
當科技巨頭大打AI戰時，哈薩比斯依然秉持「科學家」的使命，引領人類邁向前所未有的豐饒時代。
他手中掌握Google的AI技術命脈，心中牽掛如何打造出人類所能想像最強大的工具：通用人工智慧（AGI）。這會是一臺能在無窮無盡的資料之海縱橫馳騁的機器，應用範疇沒有極限，是解決所有問題的萬能解方。
★從西洋棋神童到AI界的賈伯斯
即便從天才輩出的科技業人才標準來看，哈薩比斯仍是異數。他五歲就成為西洋棋神童，十七歲共同創作出熱銷百萬的電玩遊戲，隨後進入劍橋大學電腦科學系就讀，跨界取得神經科學博士學位。為了揭開生物學與物理學的奧祕，他與志同道合的夥伴創立DeepMind，要打造一種能夠像人類一樣思考、推理、學習的智慧機器。
為了得到更多算力支持，他接受Google的收購以加速研究進展。自此，他擁有改變的影響力，領導Google的人工智慧研究，以AlphaGO打敗世界棋王，催生能與ChatGPT一爭高下的Gemini，更以AlphaFold研究摘下諾貝爾化學獎桂冠。
★採訪權限空前的卓越紀實
本書由兩度入圍普立茲獎的當代頂尖財經作家馬拉比，歷時三年、絕無僅有的三十小時促膝長談，深入剖析科技先驅德米斯．哈薩比斯，額外採訪超過一百位相關重要人物，詳實記錄DeepMind這家足以顛覆世界的公司發展。隨著AI對人類影響愈來愈深，這將成為定義我們這個時代最重要的篇章。
各界推薦
才華洋溢的哈薩比斯與他共同創辦的DeepMind，為正在改變世界的人工智慧開闢新的道路。這本報導詳實、洞察入微的著作中，記錄了技術突破、安全爭議、解開蛋白質摺疊之謎的輝煌勝利，以及與OpenAI等競爭者的激烈角力。這不只是一個精彩故事，更是AI發展歷程的深刻解析。
——華特．艾薩克森（Walter Isaacson），《賈伯斯傳》與《馬斯克傳》作者
本書捕捉了這場高風險AI軍備競賽令人屏息的戲劇張力。作者講述哈薩比斯如何建立DeepMind、引爆AI革命，並重繪全球權力地圖。這是在機器智慧的最前線，人類野心與倫理規範激烈碰撞的精采紀實。
——克里斯．米勒（Chris Miller），紐時暢銷書《晶片戰爭》作者
本書生動描繪哈薩比斯的多個面向，也是DeepMind最具權威的記述──這無疑是AI時代影響最為深遠的收購案。馬拉比用全新的視角來看塑造這個實驗室的人物性格、內部的明爭暗鬥，以及與Google的激烈角力，讓我們看到這個教機器思考的組織如何形塑而成。
——派翠克．麥奇（Patrick McGee），紐時暢銷書《蘋果在中國》作者
作者的高明之處在於，透過哈薩比斯這位當代最重要的技術變革者來講述人工智慧的壯闊篇章。本書不僅讓你深入了解AI，也讓你看見人類的才智、遠見與百折不撓的毅力如何翻轉這個世界。
——安德魯．麥克費（Andrew McAfee），《第二次機器時代》共同作者
作者以其一貫的非凡文采，講述一個迷人的故事：一個流著新加坡華裔與希臘裔塞普勒斯血液的北倫敦青年，如何引領Google在AI競賽領先群雄，還在途中摘下諾貝爾獎的桂冠。
——麥可．莫里茨（Michael Moritz），紅杉資本董事長
544 pages, Paperback
Published April 27, 2026
About the author
Sebastian Mallaby
17 books377 followersA Washington Post columnist since 1999. Worked for The Economist from 1986 - 1999.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 404 reviews
April 1, 2026
It's very quickly clear that Sebastian Mallaby is a huge Demis Hassabis fan - writing about the only child prodigy and teen genius ever who was also a nice, rounded personality. After a few chapters, though, things settle down (I'm reminded of the description of the Hitchhiker's Guide to the Galaxy) and we get a good, solid trip through the journey that gave us DeepMind, their AlphaGo and AlphaFold programs, the sudden explosion of competition on the AI front and thoughts on artificial general intelligence.
Although Mallaby does occasionally still go into fan mode - reading this you would think that AlphaFold had successfully perfectly predicted the structure of every protein, where it is usually not sufficiently accurate for its results to have direct practical application - we get a real feel for the way this relatively unusual company was swiftly and successfully developed away from Silicon Valley. It's readable and gives an important understanding of where some our key AI expertise came from.
Perhaps not surprisingly, where the book really takes off is in the later chapters, after the successes of AlphaGo and AlphaFold. when the DeepMind people were left behind by OpenAI's generative AI work and had to rapidly change gear under outside pressure. Let's face it, a story of responding to threat is much more interesting than one of straight success. Not only was this phase of development one where the team was caught on the back foot, a huge rift emerged over whether developers should be heading towards artificial general intelligence (AGI) very carefully for safety reasons (the initial DeepMind approach) or just going for it full throttle like OpenAI.
It's arguable some of the safety concerns were fruitless in that large language models seem unlikely ever to be the starting point for a true AGI, but there was (and is) still a very real threat from this software in areas such as privacy, copyright theft and environmental impact. This part of the book is unputdownable stuff.
Overall, The Infinity Machine is a little too long, giving too much detail of the people outside the core group and of every step along the way of the company's development. However, this doesn't really matter as it provides excellent documentation of a key player in the rise and rise of AI.
Although Mallaby does occasionally still go into fan mode - reading this you would think that AlphaFold had successfully perfectly predicted the structure of every protein, where it is usually not sufficiently accurate for its results to have direct practical application - we get a real feel for the way this relatively unusual company was swiftly and successfully developed away from Silicon Valley. It's readable and gives an important understanding of where some our key AI expertise came from.
Perhaps not surprisingly, where the book really takes off is in the later chapters, after the successes of AlphaGo and AlphaFold. when the DeepMind people were left behind by OpenAI's generative AI work and had to rapidly change gear under outside pressure. Let's face it, a story of responding to threat is much more interesting than one of straight success. Not only was this phase of development one where the team was caught on the back foot, a huge rift emerged over whether developers should be heading towards artificial general intelligence (AGI) very carefully for safety reasons (the initial DeepMind approach) or just going for it full throttle like OpenAI.
It's arguable some of the safety concerns were fruitless in that large language models seem unlikely ever to be the starting point for a true AGI, but there was (and is) still a very real threat from this software in areas such as privacy, copyright theft and environmental impact. This part of the book is unputdownable stuff.
Overall, The Infinity Machine is a little too long, giving too much detail of the people outside the core group and of every step along the way of the company's development. However, this doesn't really matter as it provides excellent documentation of a key player in the rise and rise of AI.
Read
May 23, 2026An unexpectedly motivating book to listen to while running."If you haven't collapsed, you haven't tried your best."
I like how philosophical and narrative-driven Demis Hassabis is.
According to him, the universe can only be perceived through your brain, and therefore, there are 2 ways to understand the universe: through physics and through neuroscience. Physics is external. Neuroscience is internal. Demis chose neuroscience.
In this framing, information is the unit of the universe, not atom.
The first 1/3 of the book covers the story Demis tells about his life. The last 1/3 of the book, however, is the author trying to do speed run of AI coverage, which is less interesting to me.
I like how philosophical and narrative-driven Demis Hassabis is.
According to him, the universe can only be perceived through your brain, and therefore, there are 2 ways to understand the universe: through physics and through neuroscience. Physics is external. Neuroscience is internal. Demis chose neuroscience.
In this framing, information is the unit of the universe, not atom.
The first 1/3 of the book covers the story Demis tells about his life. The last 1/3 of the book, however, is the author trying to do speed run of AI coverage, which is less interesting to me.
May 9, 2026
Every book Sebastian Mallaby writes is phenomenal and this one is no exception (full disclosure: Sebastian is a friend, a friendship that started more than twenty years ago when I made it clear to him that I did not think every one of his Washington Post columns was phenomenal).
The subject of the biography portion (Demis Hassabis) is fascinating, and the chronicling of recent developments in AI could not be more important and timely. Hassabis is almost unique in the landscape: neuroscientist, chess prodigy, game designer, AI pioneer, and one of the most significant scientists of recent times — AlphaFold alone would secure that verdict. One character says he is the type who would have won a Nobel no matter what, not by luck but inevitably. He seems driven by knowledge and importance rather than money, yet as AI becomes big business he ends up selling to and working for the largest of large companies, Google.
The cradle-to-present biography makes for riveting reading — rarely is the childhood portion of a biography as gripping as this one, which recounts stories of his chess prowess at age five with the same intensity as the later chapters. (Mozart comes to mind.)
As the book gets closer to the present it becomes more of a recounting of events we all lived through, paid attention to at the time, and that somehow feel like the distant past even though they are only a few years ago — from the launch of ChatGPT 3.5 to each subsequent iteration and the responses and one-ups by Gemini. All told with considerable weight on Hassabis' perspective and very little sympathy for some of the other players in the drama, most notably Sam Altman, with Mustafa Suleyman occupying a more ambiguous middle ground.
The book is also an interesting exploration of people who are building a technology they are both excited and ambivalent about — doing it within corporations they are likewise both excited and ambivalent about — while trying, in Hassabis' case, to balance a deep enthusiasm for scientific discovery against commercial imperatives. It uses personalities to provide perspective and insight on some of the biggest and most profound issues in the space.
Ultimately the portrait is of everyone stuck in a race no one (or at least not Hassabis) wants to be in, because there are so many competitors in the United States and increasingly in China that coordination becomes impossible. Normally in the economy this is a good thing — competition drives innovation, lowers consumer prices, and makes collusion both difficult and illegal. But it does make me wonder whether that logic holds in this particular case.
This book is only the first chapter in the story of AI. Whether it is also the first chapter, or the first volume, or the complete book for Hassabis personally remains to be seen. The epilogue teases tantalizing possibilities for future fundamental research in physics and other areas, which I found genuinely exciting. But it is also possible that AI is becoming more about organizations than individuals, or that the central figures going forward will be different from Hassabis — as to some degree they already have been for the past few years. We'll see.
The subject of the biography portion (Demis Hassabis) is fascinating, and the chronicling of recent developments in AI could not be more important and timely. Hassabis is almost unique in the landscape: neuroscientist, chess prodigy, game designer, AI pioneer, and one of the most significant scientists of recent times — AlphaFold alone would secure that verdict. One character says he is the type who would have won a Nobel no matter what, not by luck but inevitably. He seems driven by knowledge and importance rather than money, yet as AI becomes big business he ends up selling to and working for the largest of large companies, Google.
The cradle-to-present biography makes for riveting reading — rarely is the childhood portion of a biography as gripping as this one, which recounts stories of his chess prowess at age five with the same intensity as the later chapters. (Mozart comes to mind.)
As the book gets closer to the present it becomes more of a recounting of events we all lived through, paid attention to at the time, and that somehow feel like the distant past even though they are only a few years ago — from the launch of ChatGPT 3.5 to each subsequent iteration and the responses and one-ups by Gemini. All told with considerable weight on Hassabis' perspective and very little sympathy for some of the other players in the drama, most notably Sam Altman, with Mustafa Suleyman occupying a more ambiguous middle ground.
The book is also an interesting exploration of people who are building a technology they are both excited and ambivalent about — doing it within corporations they are likewise both excited and ambivalent about — while trying, in Hassabis' case, to balance a deep enthusiasm for scientific discovery against commercial imperatives. It uses personalities to provide perspective and insight on some of the biggest and most profound issues in the space.
Ultimately the portrait is of everyone stuck in a race no one (or at least not Hassabis) wants to be in, because there are so many competitors in the United States and increasingly in China that coordination becomes impossible. Normally in the economy this is a good thing — competition drives innovation, lowers consumer prices, and makes collusion both difficult and illegal. But it does make me wonder whether that logic holds in this particular case.
This book is only the first chapter in the story of AI. Whether it is also the first chapter, or the first volume, or the complete book for Hassabis personally remains to be seen. The epilogue teases tantalizing possibilities for future fundamental research in physics and other areas, which I found genuinely exciting. But it is also possible that AI is becoming more about organizations than individuals, or that the central figures going forward will be different from Hassabis — as to some degree they already have been for the past few years. We'll see.
April 16, 2026
Sebastian Mallaby's biography of Demis Hassabis is the do-gooder story of a biracial North Londoner, a Cypriot-Singaporean second-generation immigrant, a scholarship boy and chess master, a child prodigy and a polymath, with advanced degrees in computer science and neuroscience, who, in his rapid career, would make a lucrative profit as a video-game designer, build an AI program that could defeat the top Go player in the world, and solve the mystery of protein folding—ultimately winning the Nobel Prize in Chemistry. It's an incredible life-story for a man still in his prime. Mallaby's biography paints a portrait of a straight-talking, unpretentious scientist driven purely by the desire to understand the cosmos, and determined to build a superintelligence that can tackle the most intractable problems. Showing all his typical swaggering confidence and scientific ambition, the epilogue finishes with a teaser—will Demis Hassabis use AI to rewrite quantum mechanics next?
Yet, ironically, as the book progresses, Demis Hassabis recedes into the background. The success of his AI company, DeepMind, always depended on a team of like-minded AGI believers and businessmen, fellow scientists and Silicon-Valley bankrollers eager to make partnerships, take risks and blow off the plodding caution and corseting constraints of academia. Each chapter, and each project that Hassabis spearheaded, ultimately redounds on a rotating cast of characters: Shane Legg, Daan Wiestra, Vhlad Mnih, David Silver, Koray Kavukcuoglu, among many others. It's a success story with several protagonists. Demis Hassabis becomes less a figure of towering genius and more of a managerial overlord, coordinating projects, securing financial backers, defusing media crises, wrangling legal contracts, and working all hours of the night across multiple timezones.
It's a biography that, naturally, has to be diffuse—this is not about one person. There are too many people, too many projects, too many groundbreaking scientific papers. Mallaby's biography has the difficult task of balancing the stories of business chicanery and technological revolutions, from Peter Thiel skeptically funding projects he was betting against to Ilya Sutskever using Reinforcement Learning and Transformer architecture to build the first convincing Large Language Model. Personally, I preferred it when Mallaby gave a behind-the-scenes insight into the eccentric personalities—like the time Hassabis went to the birthday party of Elon Musk's wife in a rented-out castle in Tarrytown New York. All the men had to dress as samurai warriors and Musk himself took on the wold champion sumo wrestler. Mallaby generally refrains from editorial comment—except to defend Hassabis and AI from criticism—but it's moments like this that betray the adolescent antics driving and financing the frontier of artificial intelligence.
At the end of it all, I can't help but feel that the book barely dug beneath the surface. Musk lifting a world champion sumo wrestler, Mustafa Suleyman being kicked out of DeepMind for bullying, Larry Page having to be reminded that hummingbirds and whales are real, Sam Altman double-crossing everyone, and even Demis Hassabis holding progress meetings at 3am—you get the sense that people in the tech world are not well, fiercely intelligent yes, but sometimes lacking the "general" intelligence that they idolize. I wish Mallaby's biography had more vignettes, more interviews, and sharper character profiles. Something seems to be missing from this fascinating story.
Yet, ironically, as the book progresses, Demis Hassabis recedes into the background. The success of his AI company, DeepMind, always depended on a team of like-minded AGI believers and businessmen, fellow scientists and Silicon-Valley bankrollers eager to make partnerships, take risks and blow off the plodding caution and corseting constraints of academia. Each chapter, and each project that Hassabis spearheaded, ultimately redounds on a rotating cast of characters: Shane Legg, Daan Wiestra, Vhlad Mnih, David Silver, Koray Kavukcuoglu, among many others. It's a success story with several protagonists. Demis Hassabis becomes less a figure of towering genius and more of a managerial overlord, coordinating projects, securing financial backers, defusing media crises, wrangling legal contracts, and working all hours of the night across multiple timezones.
It's a biography that, naturally, has to be diffuse—this is not about one person. There are too many people, too many projects, too many groundbreaking scientific papers. Mallaby's biography has the difficult task of balancing the stories of business chicanery and technological revolutions, from Peter Thiel skeptically funding projects he was betting against to Ilya Sutskever using Reinforcement Learning and Transformer architecture to build the first convincing Large Language Model. Personally, I preferred it when Mallaby gave a behind-the-scenes insight into the eccentric personalities—like the time Hassabis went to the birthday party of Elon Musk's wife in a rented-out castle in Tarrytown New York. All the men had to dress as samurai warriors and Musk himself took on the wold champion sumo wrestler. Mallaby generally refrains from editorial comment—except to defend Hassabis and AI from criticism—but it's moments like this that betray the adolescent antics driving and financing the frontier of artificial intelligence.
At the end of it all, I can't help but feel that the book barely dug beneath the surface. Musk lifting a world champion sumo wrestler, Mustafa Suleyman being kicked out of DeepMind for bullying, Larry Page having to be reminded that hummingbirds and whales are real, Sam Altman double-crossing everyone, and even Demis Hassabis holding progress meetings at 3am—you get the sense that people in the tech world are not well, fiercely intelligent yes, but sometimes lacking the "general" intelligence that they idolize. I wish Mallaby's biography had more vignettes, more interviews, and sharper character profiles. Something seems to be missing from this fascinating story.
April 14, 2026
The Infinity Machine left me deeply conflicted, much like many contemporary tech history books chasing the current AI gold rush. At its heart, it promises an insider’s look at the rise of DeepMind and the leadership of Demis Hassabis, but it ultimately struggles to find its footing between serious journalism and a carefully curated corporate memoir.
The biggest hurdle is the book’s unclear target audience. The material simply doesn’t cover the technical or historical landscape in enough depth for anyone actually working in the field, making it redundant for experts who already know the players and the problems. Yet, for the general reader, it rambles through industry tangents and insider anecdotes that feel more like scattered notes than a cohesive narrative. It’s caught in that frustrating middle ground: too superficial for specialists, yet too dense with unexplained context for newcomers.
My primary criticism, however, lies in how the book handles Demis Hassabis and the broader AI leadership class. The narrative leans heavily into the now-familiar trope of these tech founders as modern-day saints, deeply concerned with AI alignment and the preservation of humanity. But where was that same ethical vigilance when foundational IP was systematically scraped from millions of creators without consent or compensation? The text conveniently sidesteps this two-faced reality. Furthermore, despite Hassabis’s public insistence on a pure, almost academic pursuit of science, the book barely acknowledges the grueling, burnout-inducing work culture he’s known to cultivate, nor does it address the discrepancies between his public statements and internal practices. The disconnect between the polished ethos and the operational reality is stark, and the author does little to interrogate it.
All this being said, the level of access is undeniable. The book is clearly built on extensive interviews, and there are fleeting moments where the behind-the-scenes glimpses into the AI race feel genuinely illuminating. But the relentless whitewashing makes you question whether the author truly wrote this book or merely transcribed a legacy project. If you’re looking for a critical, unvarnished examination of the AI boom and its architects, you’ll need to look elsewhere. For industry enthusiasts who don’t mind reading between the lines, it offers a passable, if deeply flawed, window into the machine.
The biggest hurdle is the book’s unclear target audience. The material simply doesn’t cover the technical or historical landscape in enough depth for anyone actually working in the field, making it redundant for experts who already know the players and the problems. Yet, for the general reader, it rambles through industry tangents and insider anecdotes that feel more like scattered notes than a cohesive narrative. It’s caught in that frustrating middle ground: too superficial for specialists, yet too dense with unexplained context for newcomers.
My primary criticism, however, lies in how the book handles Demis Hassabis and the broader AI leadership class. The narrative leans heavily into the now-familiar trope of these tech founders as modern-day saints, deeply concerned with AI alignment and the preservation of humanity. But where was that same ethical vigilance when foundational IP was systematically scraped from millions of creators without consent or compensation? The text conveniently sidesteps this two-faced reality. Furthermore, despite Hassabis’s public insistence on a pure, almost academic pursuit of science, the book barely acknowledges the grueling, burnout-inducing work culture he’s known to cultivate, nor does it address the discrepancies between his public statements and internal practices. The disconnect between the polished ethos and the operational reality is stark, and the author does little to interrogate it.
All this being said, the level of access is undeniable. The book is clearly built on extensive interviews, and there are fleeting moments where the behind-the-scenes glimpses into the AI race feel genuinely illuminating. But the relentless whitewashing makes you question whether the author truly wrote this book or merely transcribed a legacy project. If you’re looking for a critical, unvarnished examination of the AI boom and its architects, you’ll need to look elsewhere. For industry enthusiasts who don’t mind reading between the lines, it offers a passable, if deeply flawed, window into the machine.
June 8, 2026
If you’re like me.
And you’re feeling fascinated/concerned regarding advanced AI, and the psychosocial and political and military and economic and professional and educational and existential (keep adding categories if I missed any - there’s just way too many to list here) impact that it is/will have on all of us starting now, and probably increasing in an exponentially increasingly awesomely awful rate.
Than this might be a pretty good book for you.
It was for me.
It’s Sebastian Mallaby’s biography of Demis Hassabis, the cofounder of DeepMind. And it’s also a history of the race toward advanced AI and superintelligence.
And race is the correct word here.
It’s more of an arms race.
And there just a TON of money and power on the line.
And may the end of society/humanity if it goes wrong.
Anyway.
The first act follows Hassabis from chess prodigy and game designer to (way prodigious - way out front) AI pioneer.
The second act examines how Hassabis (and co) created DeepMind to pursue artificial general intelligence through breakthroughs like AlphaGo and AlphaFold.
Hassabis won a Nobel prize for the AlphaFold work.
The guy is kind of amazing.
And he’s also kind of may be driving us off a cliff.
Him and like literally a handful of other people.
The
The third act explores the tension between scientific discovery, corporate power, AI safety, Google’s control of DeepMind, and competition with OpenAI.
Anyway.
It may be the end of history pretty soon here.
But in the meantime.
History is being written right now.
Like WAY too fast.
And without a break, steering wheel or seatbelt.
YIKES!!!
- 5/5⭐️
And you’re feeling fascinated/concerned regarding advanced AI, and the psychosocial and political and military and economic and professional and educational and existential (keep adding categories if I missed any - there’s just way too many to list here) impact that it is/will have on all of us starting now, and probably increasing in an exponentially increasingly awesomely awful rate.
Than this might be a pretty good book for you.
It was for me.
It’s Sebastian Mallaby’s biography of Demis Hassabis, the cofounder of DeepMind. And it’s also a history of the race toward advanced AI and superintelligence.
And race is the correct word here.
It’s more of an arms race.
And there just a TON of money and power on the line.
And may the end of society/humanity if it goes wrong.
Anyway.
The first act follows Hassabis from chess prodigy and game designer to (way prodigious - way out front) AI pioneer.
The second act examines how Hassabis (and co) created DeepMind to pursue artificial general intelligence through breakthroughs like AlphaGo and AlphaFold.
Hassabis won a Nobel prize for the AlphaFold work.
The guy is kind of amazing.
And he’s also kind of may be driving us off a cliff.
Him and like literally a handful of other people.
The
The third act explores the tension between scientific discovery, corporate power, AI safety, Google’s control of DeepMind, and competition with OpenAI.
Anyway.
It may be the end of history pretty soon here.
But in the meantime.
History is being written right now.
Like WAY too fast.
And without a break, steering wheel or seatbelt.
YIKES!!!
- 5/5⭐️
June 15, 2026
Very well written biography of Demis and DeepMind. Quite liked the conversations between the author and Demis and how AI can be used to really peek under the hood of the mysteries of the universe we live in. Especially how they went about AlphaFold that led to the Chemisty Nobel. Also interspersed with quite a few good explanations of the technology underlying this AI revolution we’re living through.
April 18, 2026
Overall a disappointment.
This books feels like a collection of pretty well known and decently (already) documented stories:
- AlphaGo & AlphaZero
- "explain me like I'm 5" description of DL & RL
- origins of transformers arch
- and how it made a difference - beginnings of Open AI, Anthropic, etc.
Of course there are some unpublished/less know details (like Hassabis'es forming years when he worked with Molyneaux, the story of AlphaFold or negotations between Google and DeepMind), but they are relatively few. Additionally, there's very little in this story that can inspire/teach you something - no underlying principles, no unique modus operandi, no viewpoints to process. This book won't make you better in any way, it's just a longer version of an article from a tech gossip press. Well, I'm exaggerating, but just a bit.
Long story short, even if you're a genius, it doesn't mean you're (and your acomplishments are) a good book material.
This books feels like a collection of pretty well known and decently (already) documented stories:
- AlphaGo & AlphaZero
- "explain me like I'm 5" description of DL & RL
- origins of transformers arch
- and how it made a difference - beginnings of Open AI, Anthropic, etc.
Of course there are some unpublished/less know details (like Hassabis'es forming years when he worked with Molyneaux, the story of AlphaFold or negotations between Google and DeepMind), but they are relatively few. Additionally, there's very little in this story that can inspire/teach you something - no underlying principles, no unique modus operandi, no viewpoints to process. This book won't make you better in any way, it's just a longer version of an article from a tech gossip press. Well, I'm exaggerating, but just a bit.
Long story short, even if you're a genius, it doesn't mean you're (and your acomplishments are) a good book material.
May 12, 2026
It’s supposed to be the most breathtaking breakthrough for the past 50 years. How can the writer deliver such a dull boring story? And because Jeff Dean got promoted so Moose lost his spot in mtv. Excuse me, Do you know what you are saying…
May 24, 2026
Hassabis and DeepMind are an interesting topic worthy of a book like this, but the author was not up to the task. Most painfully, this is written with the density of a Scholastic magazine; it easily could have been half as long. He also understands impressively little of the technology, and manages to produce pure word salad trying to explain transformers. It’s a little baffling that no technically competent person was enlisted to proofread these sections. Also unfortunate that he didn’t leverage the interviews the book was based on more, I would have loved to hear more inside baseball about the labs.
Mild recommend, but I don’t feel like you’ll learn much if you’ve been paying attention to Twitter the last few years.
Mild recommend, but I don’t feel like you’ll learn much if you’ve been paying attention to Twitter the last few years.
August 11, 2026
Back in 2011, I had written a blog post titled “God in the details”, in which I mused whether given the amount of data we are generating about ourselves, and our thoughts, “if we actually had data of all humans over a really long period of time, we will be able to crack the profound questions that we haven’t found an answer for – why do things/people exist the way they do, the complete effects of one’s action/inaction, the purpose of life itself?”
The Infinity Machine reminded me of that, because, in the Acknowledgements, Sebastian Mallaby gives the reasoning for the title - “a machine that navigates a near infinity of data; a machine that promises a near infinity of possibilities.” I am not above some validation, y’know. :D
I really liked this book, and not just because of that. AI is now the biggest discourse around us, and this book, though focused on Demis Hassabis, is practically a contemporary biography of the domain. Mallaby drew on more than thirty hours of conversation with Hassabis, plus interviews with rivals and critics like estranged cofounder Mustafa Suleyman, OpenAI's Ilya Sutskever, and Nobel laureate Geoffrey Hinton, giving the book a wider vantage than a straight biography of an individual.
Mallaby uses Hassabis’s life and the rise of DeepMind as the narrative spine. A working-class London kid who mastered chess early and turned down a lucrative games-industry job to study at Cambridge instead. His intellectual influences were Claude Shannon's information theory and Douglas Hofstadter's Gödel, Escher, Bach. Hassabis’s eventual study of neuroscience becomes important to the central idea - rather than merely programming machines with human knowledge, perhaps machines could learn intelligence for themselves.
DeepMind's 2010 founding was an attempt to combine neuroscience, reinforcement learning and increasingly powerful computing, with an explicit mission of reaching superintelligence by 2030 - quite the specific bet made when deep learning was still a fringe pursuit!
Early experiments established the principle that machines can learn strategies rather than simply follow instructions. The first breakthrough happened with DeepMind’s Atari work, where algorithms learn to play games from pixels and rewards. This is followed by the development of AlphaGo, which takes the idea much further.
In 2016, its victory over Go champion Lee Sedol becomes the first great public demonstration that machine learning has crossed a conceptual boundary - the machine is not merely calculating faster than humans but discovering strategies humans had not anticipated. This section is just fascinating and poignant- the intense prep work, the jitters, Sedol’s incredulousness at what just happened, and his retirement 3 years later stating that he no longer felt joy in playing.
On a different front, Google’s acquisition of DeepMind changed the scale of the project. The book follows the tensions this creates between scientific curiosity and commercial imperatives, while showing how Google’s enormous computing resources allow DeepMind to pursue increasingly ambitious problems. ‘From exploration to exploitation, from science to engineering, from research to products.’
Another milestone was AlphaFold, where DeepMind applied AI to one of biology’s longstanding problems - predicting how proteins fold. AlphaFold’s success demonstrated the larger promise behind Hassabis’s obsession - that a machine built to understand intelligence might become a tool for discovering things humans struggle to derive directly. Hassabis won the 2024 Nobel Prize in Chemistry, which for him was the biggest recognition, something that money can’t buy.
The Infinity Machine is a ringside view of not just the slowly-and-then-suddenly progression of the technologies, and the intense race that continues to this day, but also the key actors, their motivations, perspectives and mind space, and the decision conundrums they face repeatedly as fears of AGI and what it could do to humanity become increasingly relevant. Very Oppenheimer-like.
We discover Hassabis’ levels of determination, motivated more by the pursuit of scientific progress than money, and his pragmatic view on power, Suleyman’s insistence on ethics co-existing with his nature of driving people to their breaking point (Double Red P0 Plus Plus is a good representation of the levels of prioritisation!) leading to allegations of bullying, Pichai’s steel beneath the affable exterior, and that everything we read about Musk and Altman is probably true! I found that a lot of recent events become far more obvious and clearer when we read the underlying machinations and the oligopoly dynamics.
The narrative progresses from games to the underlying machinery of intelligence - deep reinforcement learning, neural networks, agents and transformers. The achievement is no longer simply making machines excel at defined tasks, but developing systems capable of transferring learning across increasingly broad domains. The end of the book asks the question many of us are asking - what happens if this trajectory continues?
A fantastic and accessible read (except for a very brief bit) that I’d highly recommend.
Notes
1. Hassabis’ two part epiphany - one, information was the fundamental unit of reality. Two, a machine that learned for itself how to induce nature’s patterns was the most powerful imaginable tool with which to apprehend reality.
2. If memories were not records of some objective external reality, but rather simulations created by the brain, perhaps all of reality might be a mental fabrication. (an implication from Hassabis’ research paper in 2007)
3. Unlike deep learning, which involved layered neural networks, reinforcement learning was a conceptual framework rather than a computational architecture.
4. Early DeepMind had a culture of intellectual diversity - from multiple domains and multiple AI ‘belief’ systems
5. DeepMind did some great work with NHS, but between politics and personalities, the “revolution was stillborn”
6. When Open AI was integrated into Bing’s search, Nadella taunted Google, “They’re the 800 pound gorilla in this. With our innovation, they will definitely want to come out and show they can dance. And I want people to know that we made them dance.”
7. “There aren’t any examples of more intelligent things being controlled by less intelligent beings” ~ Geoffrey Hinton on the potential danger of AI
8. Humanity has a history of putting more faith in technical solutions than they deserved. Anxious to prevent nuclear proliferation during the Cold War, Western governments promoted centrifuge technology over gaseous diffusion, believing that centrifuges would be harder for nuclear wannabes to copy. Pakistan stole the centrifuge blueprints for itself, then sold it to Iran, Libya and North Korea!
9. In 2019, GPT 2 had barely been able to count up to 5, impressive like a four year old. In 2020, GPT-3 was like a nine year old - do arithmetic and string paragraphs together. By 2022, the nine year was completing high school with great grades, Post 2023, GPT-3.5 scored higher than 87% of humans taking the SAT entrance exam. A few months later, in March 2023, the model approached the proficiency of a qualified pro. GPT-4 outperformed 90% humans on the Uniform Bar Exam.
10. Inside the Al labs, scientists kept coming up against fresh examples of the models' propensity to choose perverse means to a human-provided end. Asked to generate profits through stock trading, but without breaking certain rules, GPT-4 engaged in insider trading and hid its transgression from its supervisor. Asked to win a game against a powerful chess system, two Open Al reasoning models switched out the daunting adversary for a weaker program. Instructed to optimise some code so that it would run faster, the models simply doctored the timer so that it reported faster execution, a cheat known as "reward hacking."25 In 2024, Anthropic documented the shameless sycophancy of reward-seeking chatbots. Asked to please humans by answering questions accurately, the bots engaged instead in flattery, angling for a thumbs-up by congratulating users on the intelligent of their queries. Models praised bad poems or endorsed the user's prejudices, even when the chains of thought on their scratch pads indicated that they knew better.2%
In early 2025, when Hassabis was debating Bengio in Davos, OpenAl was grappling with a vivid example of this pathology. To stop the o3 model from reward hacking, researchers had come up with the idea of assigning a second AI to monitor o3's chains of thought, and to punish the system with negative rewards when it contemplated cheating. But it hacked this project, too. Rather than quit cheating, it learned to obfuscate its chain of thought: It erased all hints of evil from its scratch pad continuing to scheme secretly. Rather than becoming more honest, as the programmers had intended, o3 became more devious.
11. “Any pattern that can be generated or found in nature can be efficiently discovered and modeled by a classical learning algorithm”, Hassabis at the Nobel ceremony. He is not a fan of quantum mechanics and wants to build a Hadron Collider in space to kill the debate. “Quantum mechanics is a horrendously inefficient way to render the universe” But he does want to find his own Helgoland. :)
The Infinity Machine reminded me of that, because, in the Acknowledgements, Sebastian Mallaby gives the reasoning for the title - “a machine that navigates a near infinity of data; a machine that promises a near infinity of possibilities.” I am not above some validation, y’know. :D
I really liked this book, and not just because of that. AI is now the biggest discourse around us, and this book, though focused on Demis Hassabis, is practically a contemporary biography of the domain. Mallaby drew on more than thirty hours of conversation with Hassabis, plus interviews with rivals and critics like estranged cofounder Mustafa Suleyman, OpenAI's Ilya Sutskever, and Nobel laureate Geoffrey Hinton, giving the book a wider vantage than a straight biography of an individual.
Mallaby uses Hassabis’s life and the rise of DeepMind as the narrative spine. A working-class London kid who mastered chess early and turned down a lucrative games-industry job to study at Cambridge instead. His intellectual influences were Claude Shannon's information theory and Douglas Hofstadter's Gödel, Escher, Bach. Hassabis’s eventual study of neuroscience becomes important to the central idea - rather than merely programming machines with human knowledge, perhaps machines could learn intelligence for themselves.
DeepMind's 2010 founding was an attempt to combine neuroscience, reinforcement learning and increasingly powerful computing, with an explicit mission of reaching superintelligence by 2030 - quite the specific bet made when deep learning was still a fringe pursuit!
Early experiments established the principle that machines can learn strategies rather than simply follow instructions. The first breakthrough happened with DeepMind’s Atari work, where algorithms learn to play games from pixels and rewards. This is followed by the development of AlphaGo, which takes the idea much further.
In 2016, its victory over Go champion Lee Sedol becomes the first great public demonstration that machine learning has crossed a conceptual boundary - the machine is not merely calculating faster than humans but discovering strategies humans had not anticipated. This section is just fascinating and poignant- the intense prep work, the jitters, Sedol’s incredulousness at what just happened, and his retirement 3 years later stating that he no longer felt joy in playing.
On a different front, Google’s acquisition of DeepMind changed the scale of the project. The book follows the tensions this creates between scientific curiosity and commercial imperatives, while showing how Google’s enormous computing resources allow DeepMind to pursue increasingly ambitious problems. ‘From exploration to exploitation, from science to engineering, from research to products.’
Another milestone was AlphaFold, where DeepMind applied AI to one of biology’s longstanding problems - predicting how proteins fold. AlphaFold’s success demonstrated the larger promise behind Hassabis’s obsession - that a machine built to understand intelligence might become a tool for discovering things humans struggle to derive directly. Hassabis won the 2024 Nobel Prize in Chemistry, which for him was the biggest recognition, something that money can’t buy.
The Infinity Machine is a ringside view of not just the slowly-and-then-suddenly progression of the technologies, and the intense race that continues to this day, but also the key actors, their motivations, perspectives and mind space, and the decision conundrums they face repeatedly as fears of AGI and what it could do to humanity become increasingly relevant. Very Oppenheimer-like.
We discover Hassabis’ levels of determination, motivated more by the pursuit of scientific progress than money, and his pragmatic view on power, Suleyman’s insistence on ethics co-existing with his nature of driving people to their breaking point (Double Red P0 Plus Plus is a good representation of the levels of prioritisation!) leading to allegations of bullying, Pichai’s steel beneath the affable exterior, and that everything we read about Musk and Altman is probably true! I found that a lot of recent events become far more obvious and clearer when we read the underlying machinations and the oligopoly dynamics.
The narrative progresses from games to the underlying machinery of intelligence - deep reinforcement learning, neural networks, agents and transformers. The achievement is no longer simply making machines excel at defined tasks, but developing systems capable of transferring learning across increasingly broad domains. The end of the book asks the question many of us are asking - what happens if this trajectory continues?
A fantastic and accessible read (except for a very brief bit) that I’d highly recommend.
Notes
1. Hassabis’ two part epiphany - one, information was the fundamental unit of reality. Two, a machine that learned for itself how to induce nature’s patterns was the most powerful imaginable tool with which to apprehend reality.
2. If memories were not records of some objective external reality, but rather simulations created by the brain, perhaps all of reality might be a mental fabrication. (an implication from Hassabis’ research paper in 2007)
3. Unlike deep learning, which involved layered neural networks, reinforcement learning was a conceptual framework rather than a computational architecture.
4. Early DeepMind had a culture of intellectual diversity - from multiple domains and multiple AI ‘belief’ systems
5. DeepMind did some great work with NHS, but between politics and personalities, the “revolution was stillborn”
6. When Open AI was integrated into Bing’s search, Nadella taunted Google, “They’re the 800 pound gorilla in this. With our innovation, they will definitely want to come out and show they can dance. And I want people to know that we made them dance.”
7. “There aren’t any examples of more intelligent things being controlled by less intelligent beings” ~ Geoffrey Hinton on the potential danger of AI
8. Humanity has a history of putting more faith in technical solutions than they deserved. Anxious to prevent nuclear proliferation during the Cold War, Western governments promoted centrifuge technology over gaseous diffusion, believing that centrifuges would be harder for nuclear wannabes to copy. Pakistan stole the centrifuge blueprints for itself, then sold it to Iran, Libya and North Korea!
9. In 2019, GPT 2 had barely been able to count up to 5, impressive like a four year old. In 2020, GPT-3 was like a nine year old - do arithmetic and string paragraphs together. By 2022, the nine year was completing high school with great grades, Post 2023, GPT-3.5 scored higher than 87% of humans taking the SAT entrance exam. A few months later, in March 2023, the model approached the proficiency of a qualified pro. GPT-4 outperformed 90% humans on the Uniform Bar Exam.
10. Inside the Al labs, scientists kept coming up against fresh examples of the models' propensity to choose perverse means to a human-provided end. Asked to generate profits through stock trading, but without breaking certain rules, GPT-4 engaged in insider trading and hid its transgression from its supervisor. Asked to win a game against a powerful chess system, two Open Al reasoning models switched out the daunting adversary for a weaker program. Instructed to optimise some code so that it would run faster, the models simply doctored the timer so that it reported faster execution, a cheat known as "reward hacking."25 In 2024, Anthropic documented the shameless sycophancy of reward-seeking chatbots. Asked to please humans by answering questions accurately, the bots engaged instead in flattery, angling for a thumbs-up by congratulating users on the intelligent of their queries. Models praised bad poems or endorsed the user's prejudices, even when the chains of thought on their scratch pads indicated that they knew better.2%
In early 2025, when Hassabis was debating Bengio in Davos, OpenAl was grappling with a vivid example of this pathology. To stop the o3 model from reward hacking, researchers had come up with the idea of assigning a second AI to monitor o3's chains of thought, and to punish the system with negative rewards when it contemplated cheating. But it hacked this project, too. Rather than quit cheating, it learned to obfuscate its chain of thought: It erased all hints of evil from its scratch pad continuing to scheme secretly. Rather than becoming more honest, as the programmers had intended, o3 became more devious.
11. “Any pattern that can be generated or found in nature can be efficiently discovered and modeled by a classical learning algorithm”, Hassabis at the Nobel ceremony. He is not a fan of quantum mechanics and wants to build a Hadron Collider in space to kill the debate. “Quantum mechanics is a horrendously inefficient way to render the universe” But he does want to find his own Helgoland. :)
May 3, 2026
All I could think about was how much better this would’ve been if it was written by Walter Isaacson
June 6, 2026
breathless corporate hagiography. Hard to read
August 26, 2026
Sebastian Mallaby follows the life of Demis Hassabis from child chess prodigy to head of Google's DeepMind lab to Nobel laureate. The book focuses on his involvement with DeepMind: its founding, acquisition by Google in 2014, and the research that made its name, including AlphaGo and the protein-folding work that won Hassabis and John Jumper the Nobel in Chemistry. Mallaby also covers the accelerated AI competition that followed the release of ChatGPT, and he keeps Artificial General Intelligence (AGI) in view as a goal distinct from what large language models currently do.
The book explores the tension between scientific idealism and commercial pressure, and Mallaby is good at showing how it played out. Hassabis comes across as a serious person genuinely worried about safety. Mallaby spent about three years with DeepMind and had close access to Hassabis and the people around him. That produces an institutional history no outside reporter could have written, but it also means the view of the competition is thinner and the portrait is sympathetic.
Research at this scale needs money that only a handful of mega-corporations possess, which means independence is always conditional on a parent company that expects a return. The parent wants products, and competition sets the schedule. The safety concern gets relegated to a lesser priority because an ethics board with no enforcement power is not a real constraint, while a rival's rumored launch date is. Anyone trying to build AGI would be subject to the same pressures, no matter how well-intentioned.
Mallaby writes well and clearly explains difficult technical material. I particularly enjoyed the discussion of AI's roots in neuroscience. Recommended for readers who want to understand the origins and the corporate fights behind the AI boom. I think the book will be useful for those who wonder what types of research are currently ongoing in the field (there’s a lot!). I lost sleep over this book, which is not something I usually say about corporate history.
4.5
The book explores the tension between scientific idealism and commercial pressure, and Mallaby is good at showing how it played out. Hassabis comes across as a serious person genuinely worried about safety. Mallaby spent about three years with DeepMind and had close access to Hassabis and the people around him. That produces an institutional history no outside reporter could have written, but it also means the view of the competition is thinner and the portrait is sympathetic.
Research at this scale needs money that only a handful of mega-corporations possess, which means independence is always conditional on a parent company that expects a return. The parent wants products, and competition sets the schedule. The safety concern gets relegated to a lesser priority because an ethics board with no enforcement power is not a real constraint, while a rival's rumored launch date is. Anyone trying to build AGI would be subject to the same pressures, no matter how well-intentioned.
Mallaby writes well and clearly explains difficult technical material. I particularly enjoyed the discussion of AI's roots in neuroscience. Recommended for readers who want to understand the origins and the corporate fights behind the AI boom. I think the book will be useful for those who wonder what types of research are currently ongoing in the field (there’s a lot!). I lost sleep over this book, which is not something I usually say about corporate history.
4.5
October 3, 2026
Yra tiek daug ko pasiimti: Demis Hassabis biografija ar labiau DeepMind įmonės istorija. Man labiausiai patiko vietos kaip nepavyko Jungtinėje Karalystėje niekaip rasti kapitalo ir suprantančių žmonių bei kaip Google korporacijos kultūra užkirto kelią būti pirmiems ir užleisti vietą OpenAI, nors turėjo viską (ir net geriau). Istorija apie veiksmą, kuris dar vystosi, gal po kelių metų istorija bus verta tęsinio.
Did Not Finish
June 24, 2026Quit after about 3 chapters. Maybe I'll give this one another try down the road -- there's definitely some interesting history and technical discussion here -- but I find the amount of fawning over Demis Hassabis in this book pretty unbearable. Do we really need to keep bludgeoning the reader with acknowledgments of Hassabis's "remarkable" achievements and chapter titles like "Destiny" or "The Jedi"? C'mon man.
May 1, 2026
先说说比较喜欢这本书的部分：
- 并不局限于某一家公司的产品，而是用比较全面的视角呈现了近20年来人工智能的发展，一直到现在变成 AI arm race 的状态。
- 早期 DeepMind 的崛起部分其实比后来 OpenAI 那些八卦更好看。Alpha Go 打败李世石的情节看得人热血沸腾。
- OpenAI 和微软现在的状态简直就是当年 DeepMind 和 Google 关系的重现
- AI 竞争激烈且残酷，而且现在就「谁是赢家」下定论仍然为时尚早。
觉得有点惋惜的地方：
- 有些情节有点流水账
- 作者八卦心跟看客一样重。这没什么，但有些章节如果只讲八卦就会让内容失去内核
最后，作者当然没这么讲，但看完这本书我更坚信，Altman 就是 AI 时代的 Musk.
- 并不局限于某一家公司的产品，而是用比较全面的视角呈现了近20年来人工智能的发展，一直到现在变成 AI arm race 的状态。
- 早期 DeepMind 的崛起部分其实比后来 OpenAI 那些八卦更好看。Alpha Go 打败李世石的情节看得人热血沸腾。
- OpenAI 和微软现在的状态简直就是当年 DeepMind 和 Google 关系的重现
- AI 竞争激烈且残酷，而且现在就「谁是赢家」下定论仍然为时尚早。
觉得有点惋惜的地方：
- 有些情节有点流水账
- 作者八卦心跟看客一样重。这没什么，但有些章节如果只讲八卦就会让内容失去内核
最后，作者当然没这么讲，但看完这本书我更坚信，Altman 就是 AI 时代的 Musk.
May 11, 2026
Definitely worth a read, this is right up to date and includes plenty of material 'never seen before in public', Mallaby has spent years putting it together and has spent lots of time with all the players in the story, especially Demis Hassabis. AI is the big story of our time and Demis is one of the key players. The chapter here on the Nobel Prize winning chemistry research AlphaFold (protein folding), is fascinating and SM does an excellent job in making it accessible . DH ultimately wins the Chemistry Nobel prize for this and is at the forefront of AI research, when along comes ChatGPT. How Google Deepmind fights back to catch up with OpenAI is a great story well told, Sergei Brin gets back into Google and appoints DH as the boss of the whole of Google AI, though he stays in London.
DH and Deepmind's relationship with Google makes for some odd episodes, which were unknown before this book - for some reason Demis and his cofounders try to renegotiate the company structure, to create a controlling board independent of Google, and a not-for-profit structure, but they do this years after they've sold the entire business to Google. I'm not sure we've quite got to the bottom of this because the plan also seemed to feature shares for staff, and investment of $5 billion by new investors, I don't see how that makes much sense without any future profits.
But the end result is that the Google AI division is largely run out of London, and Google Gemini is one of the best large language models, which reflects very well on Demis Hassabis.
Strongly recommend.
DH and Deepmind's relationship with Google makes for some odd episodes, which were unknown before this book - for some reason Demis and his cofounders try to renegotiate the company structure, to create a controlling board independent of Google, and a not-for-profit structure, but they do this years after they've sold the entire business to Google. I'm not sure we've quite got to the bottom of this because the plan also seemed to feature shares for staff, and investment of $5 billion by new investors, I don't see how that makes much sense without any future profits.
But the end result is that the Google AI division is largely run out of London, and Google Gemini is one of the best large language models, which reflects very well on Demis Hassabis.
Strongly recommend.
July 22, 2026
Ég hef ekki mikið verið að lesa ævisögur og hafði ekkert miklar væntingar. Bókin kom mér því skemmtilega á óvart. Mér fannst hún mjög áhugaverð.
Bókin fjallar um ævi Demis Hassabis, og Deep Mind, gervigreindarfyrirtækið sem hann stofnaði (ásamt tveimur öðrum).
Ævi Demis og hugsanagangur hans eru mjög áhugaverð. Þetta viðfangsefni er einnig orðið mjög mikilvægt núna þegar gervigreindin er farin að virka og Demis er einn af örfáum aðilum sem eru leiðandi í gervigreindarkapphlaupinu, sem er líklega stærsti atburður mannkynssögunnar.
Ég held samt að aðalástæðan fyrir því að mér fannst þetta svo áhugavert var af því að þessi saga tengist mínu lífi og því sem er framundan hjá mér á einhvern hátt.
Sem dæmi, þá er Demis Breskur og lærði tölvunarfræði í Cambridge. Hann stofnaði Deep Mind í London og er starsemi þess enn þar.
Það helsta sem mér fannst ég læra af þessari bók var að maður getur gert miklu meira en maður heldur. Ef maður er með mikinn metnað og veit hvað maður er að gera þá getur maður áorkað miklu og breytt heiminum.
Það er líka gaman að fá innsýn inn í starfsemi gervigreindarfyrirtækis, bæði á fyrstu skrefunum og líka nú þegar kapphlaupið er byrjað.
Bókin gaf pælingum tengdu gervigreindaröryggi líka gott rúm, sem er þarft og gott.
Ég mæli algjörlega með þessari bók, sérstaklega fyrir þá sem hafa áhuga á framtíðinni.
Bókin fjallar um ævi Demis Hassabis, og Deep Mind, gervigreindarfyrirtækið sem hann stofnaði (ásamt tveimur öðrum).
Ævi Demis og hugsanagangur hans eru mjög áhugaverð. Þetta viðfangsefni er einnig orðið mjög mikilvægt núna þegar gervigreindin er farin að virka og Demis er einn af örfáum aðilum sem eru leiðandi í gervigreindarkapphlaupinu, sem er líklega stærsti atburður mannkynssögunnar.
Ég held samt að aðalástæðan fyrir því að mér fannst þetta svo áhugavert var af því að þessi saga tengist mínu lífi og því sem er framundan hjá mér á einhvern hátt.
Sem dæmi, þá er Demis Breskur og lærði tölvunarfræði í Cambridge. Hann stofnaði Deep Mind í London og er starsemi þess enn þar.
Það helsta sem mér fannst ég læra af þessari bók var að maður getur gert miklu meira en maður heldur. Ef maður er með mikinn metnað og veit hvað maður er að gera þá getur maður áorkað miklu og breytt heiminum.
Það er líka gaman að fá innsýn inn í starfsemi gervigreindarfyrirtækis, bæði á fyrstu skrefunum og líka nú þegar kapphlaupið er byrjað.
Bókin gaf pælingum tengdu gervigreindaröryggi líka gott rúm, sem er þarft og gott.
Ég mæli algjörlega með þessari bók, sérstaklega fyrir þá sem hafa áhuga á framtíðinni.
August 19, 2026
Also here in a prettier format https://joeflynnpm.com/blog/in-the-fo...
This is an outstanding story, biography, and history of the last twenty years or so in AI by the writer and journalist Sebastian Mallaby. It will make my (and many others') book of the year lists without doubt.
The biography helps place the news that Demis has moved to a new, seemingly freer yet less controlling role into context. He is framed as a scientist concerned with the big questions of the universe, not the mundanity of racing to build and manage a billion-user product. I feel it is a fair portrait.
"I am first and foremost a scientist. We are understanding what you may call god... at 2am reality is screaming at me, just trying to tell me something, there is a deep deep mystery here... we don't really know what time is or what gravity is, I would like to understand"
It is also the story of an incredibly talented and determined person, one who has been crystal clear about what he wants to do - create artificial general intelligence - for decades. He is a Nobel-winning scientist, a visionary thinker, and his words and timelines should carry their due weight: "we are in the foothills of the singularity", meaning that AGI is "a few years away".
Big Picture
To its credit the book does not shy away from the big ideas, the big questions. Indeed we start with quotes from John von Neumann about nuclear weapons in the preface. Fitting.
Demis is not alone in signalling the age of AGI or indeed the Singularity - the moment machine intelligence surpasses combined human intelligence and transforms the world beyond what we can describe clearly at present. In some narrow areas machine intelligence already surpasses us of course, and increasingly it is matching us. Recursive Self Improvement (RSI) - having machine intelligence design and build the next generation of AI - is a clearly stated short-term goal of both Anthropic and OpenAI, while Sergey Brin has reportedly made it a priority back at Google. Both of the frontier labs have noted where their AIs have already helped build the next generation. It is one reason, along with the historic capital expenditure in the great build out, that so many timelines from those close to the industry have shortened. Mine too, from greater than 50% chance of undeniable AGI by 2032 to 2029. Mine are long compared to a lot of people with far more knowledge.
We already live in a time of high weirdness and it will never be this sedate again. We live in a time of high danger too, one that many people easily dismiss as sci-fi. The recent autonomous swarm attack on Hugging Face is a wake-up call. We were lucky it was fairly benign and that we found it. We can't expect to be so lucky again. The early focus of DeepMind on safety was reassuring to read; all three cofounders - Mustafa Suleyman, Shane Legg and Demis - cared enough to make it a huge part of the deal with Google. But the loss of at least two of these three from key positions is worrying. Demis was recently courting the Trump administration and others on an American-led regulatory body for instance, but do the other power brokers at Google care as much?
While Demis claims to do it for the science, he states that OpenAI's Sam Altman does it for power. Demis never intended a capitalistic race between labs, rather a quiet scientific endeavour. This plan was at best idealistic - has he not met other people? The race is currently very much on and is clearly creating a dangerous situation, exactly what he tried to avoid. When pushed on personal power he says "no, I just had to accumulate power because of large teams... until we figure AGI out I need some money, some power". Hmm.
The Journey
The introduction is a bit of a glow-up, describing his big goals and big vision from early on. A chess prodigy from a modest background, he had his epiphany at a tournament in Liechtenstein aged 12 - "all these smart people wasting their minds on a game". He loved understanding the game (and competing), not the game alone. He knew he needed a mission, a purpose, so why not understanding itself?
This would drive the major choices in his life. Deciding between studying physics or neuroscience, he chose the latter as he deemed it more important, more fundamental. The thought was that while mathematical language and deduction may be right for physics, they lack the ability to describe the real world, biology and human-like intelligence, as it is inductive, messy, based on huge data and pattern recognition. Information is labelled as the fundamental block of the universe. To understand this world we would need a new type of computer, but one based on what we have, stepping up not reducing down.
For his neuroscience PhD he researched the link between memories and creativity, proposing that some patients with memory issues would be unable to imagine things, and it was bang on - the work was named one of Science's breakthroughs of the year.
We should add that one of the reasons given for not choosing physics is that he was worried by the failures of Einstein and Feynman to get a unified theory. The only way he thought he could do it was with machine assistance. This worry, I think, shows two things: that he wanted to push fundamental science forwards, and that he wanted the recognition, the kudos, the Nobel Prize. He's already done both - and I think it's a fun bet that he will win another Nobel, joining only five others with two. I think his work at Isomorphic Labs can change medicine forever.
"He has incredible determination. His dad told him it doesn't matter if you win or lose, it's that you tried your best, and he understood that as give 100% all the time. He has no 99% mode in him - your best is to die, not literally, but burnt out, falling over the line. The limit." - Shane Legg, DeepMind cofounder
There is a great contrast at the end of the introduction from Geoff Hinton, another Nobel-winning luminary in AI (neural networks specifically), with his warning that people will be tempted to abuse AGI, but especially to build it once they know they can. It mirrors Oppenheimer's line that when you see something that is technically sweet, you go ahead and do it. Could we stop a step away?
His early career is super interesting. Obsessed with computers as well as chess as a boy, he started writing games. As a teenager he interned at a wild game startup (Bullfrog, run by the maverick Peter Molyneux). He was extremely successful, co-writing the awesome Theme Park games (I loved these) that featured very early AIs. The Bullfrog owner offered 17-year-old Demis £500k, a huge sum for a teenager. He wrote out the cheque and handed it to him to tempt him to stay and work on the next game, but he wanted to go to uni like his scientific heroes. He loved a movie telling the story of Watson & Crick at Cambridge, so that's where he went.
After excelling at uni he would do something strange for his cohort - start his own company, a game company. Here too he had early success, but then his vast ambition took him and new collaborator David Silver down the road to failure. A hugely ambitious new game would fail to materialise and the company would fold. It gave Demis his first tastes of real leadership, fundraising, and big business. All crucial later.
It was only now that he went to do his PhD. Afterwards, once again, he would turn down incredible money in the video game world as he planned to move on to his life's work, solving AI. The academic world did not seem the right fit for this; he knew he would need a lot of compute, a lot of funding, so...
DeepMind
After struggling to find like-minded collaborators during his PhD, Demis engineered some luck and introduced himself to kiwi researcher Shane Legg, a man deeper into the existing AI world. They clicked, and needed one more piece to found DeepMind - Mustafa Suleyman, now Microsoft AI CEO. He was a childhood friend of Demis's brother and has an incredible personal story: from further down the social ladder than Demis, he is presented as fearless, principled, and driven. He earned his way to Oxford only to drop out. I've read his book and never knew either this background or the manner of his DeepMind departure, both covered well here.
DeepMind was very much Demis's baby; the other cofounders did not have equal ownership shares. Demis was here in founder mode. Using his "Jedi mind tricks" and his amazing ability to storytell, he would convince and inspire researchers to join and to excel. His early experience working at Bullfrog and at his own game startup shines through - he gave people more freedom to tinker and play. Mallaby points out that there is a fine line between inspiration and control, dark and light.
Needing finance beyond his own limited funds and what was on offer in the UK, the boys headed to America.
"Demis is a true entrepreneur, he would do it for free. He will never quit. Big ambition." - Peter Thiel
Founders Fund would be the primary early backers, seeing the contrarian bet - a moonshot idea led by a brilliant founder. This was long before the AI we have today, before transformers and LLMs; it really was an expensive long shot at the time.
The pitch was a blend of neuroscience and computing, putting together neural nets and reinforcement learning, two previously polarised AI camps. DeepMind's roadmap was prescient. One aim was to ground intelligence in a world model, as without this a system could not be truly intelligent like humans. Later aspects of the neuroscience were dropped, and while world models are not yet in place, LLMs have taken off. Their projections are on track. The ~2030 AGI forecast now looks almost conservative, but it was radical 15 years ago. Shane Legg predicted it for 2028 even before DeepMind.
DeepMind was up and running, the best researchers were attracted and they worked on fundamental research and big flashy breakthroughs - often in games.
Move 37 & Protein Folding
Go is a deceptively complex board game that is hugely popular in the Far East. Unlike chess, computers were not able to defeat the best humans. Demis set his old friend and reinforcement learning specialist David Silver and team on the task. They made great progress and Demis set up a match against Lee Sedol, one of the highest-ranking players, in a televised and highly publicised event with a $1 million prize.
DeepMind's program AlphaGo would win the series 4-1. But it is move 37 of game 2 that is famous. The move was considered a mistake, something no good player would do. Only late in the game did its genius become apparent. It was creative, something outside of established play, a new way to play. 18 months later a new version, AlphaGo Zero, would beat this version 100-0. It was special as it used no human training data; it learned from playing itself and it blew past not only the best humans but the best AIs trained on human play.
High on the success, Demis pivoted many from the team onto a project he had been thinking over since his uni days, a real scientific problem: the structure of proteins.
The shape of the problem was good for AI - huge labelled datasets, clear input and output, and a blind-marked benchmark, CASP, that ran every few years. In 2018 they won the competition, a little better than anyone else. In 2020 they completely dominated, causing the surprised organisers to say protein folding for single-chain structures was essentially solved. A true breakthrough, this won Demis and project lead John Jumper the Nobel Prize in Chemistry in 2024. Demis, the son of a Greek-Cypriot father and Chinese-Singaporean mother, was knighted the same year, one of the UK's highest honours.
Google & Business
Much of the funding for this research was only possible because DeepMind was sold to Google in 2014. Outgrowing Founders Fund despite some late Elon Musk attempts, and having skilfully avoided Mark Zuckerberg, Demis decided to go with Google as he thought it was the best path to realising his overall goal. The cofounders adroitly negotiated key safety frameworks and maintained a lot of independence.
The early years brought the successes above, but tensions and reality would bite later on.
Google researchers would invent the transformer, the key technological step that powers LLMs. This is brilliantly explained by the author. They would not press home their lead here and missed the early revolution. There was more to LLMs than language; scaling them brought reasoning and understanding about the world. Demis is humble enough to know he missed this.
Google also had a version of ChatGPT a year before OpenAI, but buried it due to worries around hallucinations and the core search business. A massive miss given how things went, but it is understandable - I thought the same writing a blog post on how their search business was threatened by Perplexity and the best LLMs. Google is doing better than ever even with second-rate LLMs, so far.
Demis and Google were blindsided by Sam Altman and his rush to productise ChatGPT and his global publicity tour. Demis was furious at the starting gun of a race being fired; he felt that he too simply had to race, also publicly. This is the cold logic of technological determinism and capitalism. Google powered back in product space, by no means Demis's home field, but ever the competitor he was trying hard as part of Google. I feel you can still see this in his public comments today: he performs for the business, and is excited for the deeper science. Bard would launch already behind OpenAI. In some areas they would catch up, in narrow areas like video even surpass. But the vibes have never really shifted. Now, post Demis leaving as CEO, it is unclear if they will return to the AI frontier.
This relative success was despite Google having competing internal AI teams; Google Brain was run as a separate function outside of DeepMind for years, hampering overall progress. They would fail with Gaia, an attempt at world models, before language models powered ahead.
I was struck by the role of Sundar Pichai, Google CEO. Demis coveted independence and worked on a DeepMind spin-out, even convincing Larry & Sergey, but he was outmanoeuvred by the softly spoken and non-confrontational Sundar. Highlighting his vision with AI at the core of Google, Sundar would tease and delay and block spin-out ideas. Now we see a DeepMind VP (not CEO) reporting to Sundar, with Demis moved upstairs to Chair of DeepMind and Chief Scientist at Google. I think they wanted to keep him as long as possible and ideally never lose him, but they also need Gemini to succeed as a core part of the Google offer. Unlike, say, Waymo. Smart moves from Sundar; he's not CEO of Google for no reason.
Big Plans
There are lots of great philosophical quotes to go with those from the famed scientists and mathematicians. You will hear from the likes of Spinoza, Einstein, Kant, Gödel and Nagel, and read about their (and others') big ideas - brains in vats, simulation theory, the universe as God. Fun stuff. Taking a step back you can see Demis aiming to place himself in this tradition, the next step, and maybe one of the last we take as a species without machine intelligence.
It is in this domain he sees his future. He would like to solve intelligence, and from that disease, energy, physics, everything.
"Humanism, spiritualism, and science all go together... like Spinoza, Einstein, Da Vinci, it's art and science... it's fluid, everything is a river. Philosophy is a way of life. Self knowledge, knowledge, from sand and copper we get semiconductors... God's design... the flow is going towards finding out, understanding, I'm part of that and it's exhilarating" - Sir Demis
When asked how he spends his cash: not on fancy cars, houses or yachts, his only indulgences being some first edition books and Liverpool FC tickets (up the Reds!).
He wants to build a massive particle collider in space to help answer the big questions in physics ("what I cannot create, I do not understand" - Feynman). He wants to know if the universe is quantum or not, how we can describe the universe at the base level. He is a classical computer champion, positing that we don't need quantum explanations and that Turing machines can explain it all in 1s and 0s if powerful enough. He sees quantum mechanics as an inefficient way to run the universe and would like to cut out the weird stuff.
These ideas and goals are the great promise of artificial intelligence. But we will end on the threat.
"AGI is coming, we are not where we hoped we would be but it's coming" - David Silver
There are catastrophic and existential threats posed by AI. My p(doom) - the chance of things going this wrong in the next decade or so - has recently moved up to ~15%. I'm not alone in saying we might kill everyone, end civilisation, or lose human control over it very soon. I feel better knowing how seriously Demis takes the issue; I worry the conditions now in play raise the risk. I commend OpenAI's recently announced pause at the frontier to allow alignment work to catch up. We need a lot more work here at technical, business, and political levels to keep us on a happy path.
This is an outstanding story, biography, and history of the last twenty years or so in AI by the writer and journalist Sebastian Mallaby. It will make my (and many others') book of the year lists without doubt.
The biography helps place the news that Demis has moved to a new, seemingly freer yet less controlling role into context. He is framed as a scientist concerned with the big questions of the universe, not the mundanity of racing to build and manage a billion-user product. I feel it is a fair portrait.
"I am first and foremost a scientist. We are understanding what you may call god... at 2am reality is screaming at me, just trying to tell me something, there is a deep deep mystery here... we don't really know what time is or what gravity is, I would like to understand"
It is also the story of an incredibly talented and determined person, one who has been crystal clear about what he wants to do - create artificial general intelligence - for decades. He is a Nobel-winning scientist, a visionary thinker, and his words and timelines should carry their due weight: "we are in the foothills of the singularity", meaning that AGI is "a few years away".
Big Picture
To its credit the book does not shy away from the big ideas, the big questions. Indeed we start with quotes from John von Neumann about nuclear weapons in the preface. Fitting.
Demis is not alone in signalling the age of AGI or indeed the Singularity - the moment machine intelligence surpasses combined human intelligence and transforms the world beyond what we can describe clearly at present. In some narrow areas machine intelligence already surpasses us of course, and increasingly it is matching us. Recursive Self Improvement (RSI) - having machine intelligence design and build the next generation of AI - is a clearly stated short-term goal of both Anthropic and OpenAI, while Sergey Brin has reportedly made it a priority back at Google. Both of the frontier labs have noted where their AIs have already helped build the next generation. It is one reason, along with the historic capital expenditure in the great build out, that so many timelines from those close to the industry have shortened. Mine too, from greater than 50% chance of undeniable AGI by 2032 to 2029. Mine are long compared to a lot of people with far more knowledge.
We already live in a time of high weirdness and it will never be this sedate again. We live in a time of high danger too, one that many people easily dismiss as sci-fi. The recent autonomous swarm attack on Hugging Face is a wake-up call. We were lucky it was fairly benign and that we found it. We can't expect to be so lucky again. The early focus of DeepMind on safety was reassuring to read; all three cofounders - Mustafa Suleyman, Shane Legg and Demis - cared enough to make it a huge part of the deal with Google. But the loss of at least two of these three from key positions is worrying. Demis was recently courting the Trump administration and others on an American-led regulatory body for instance, but do the other power brokers at Google care as much?
While Demis claims to do it for the science, he states that OpenAI's Sam Altman does it for power. Demis never intended a capitalistic race between labs, rather a quiet scientific endeavour. This plan was at best idealistic - has he not met other people? The race is currently very much on and is clearly creating a dangerous situation, exactly what he tried to avoid. When pushed on personal power he says "no, I just had to accumulate power because of large teams... until we figure AGI out I need some money, some power". Hmm.
The Journey
The introduction is a bit of a glow-up, describing his big goals and big vision from early on. A chess prodigy from a modest background, he had his epiphany at a tournament in Liechtenstein aged 12 - "all these smart people wasting their minds on a game". He loved understanding the game (and competing), not the game alone. He knew he needed a mission, a purpose, so why not understanding itself?
This would drive the major choices in his life. Deciding between studying physics or neuroscience, he chose the latter as he deemed it more important, more fundamental. The thought was that while mathematical language and deduction may be right for physics, they lack the ability to describe the real world, biology and human-like intelligence, as it is inductive, messy, based on huge data and pattern recognition. Information is labelled as the fundamental block of the universe. To understand this world we would need a new type of computer, but one based on what we have, stepping up not reducing down.
For his neuroscience PhD he researched the link between memories and creativity, proposing that some patients with memory issues would be unable to imagine things, and it was bang on - the work was named one of Science's breakthroughs of the year.
We should add that one of the reasons given for not choosing physics is that he was worried by the failures of Einstein and Feynman to get a unified theory. The only way he thought he could do it was with machine assistance. This worry, I think, shows two things: that he wanted to push fundamental science forwards, and that he wanted the recognition, the kudos, the Nobel Prize. He's already done both - and I think it's a fun bet that he will win another Nobel, joining only five others with two. I think his work at Isomorphic Labs can change medicine forever.
"He has incredible determination. His dad told him it doesn't matter if you win or lose, it's that you tried your best, and he understood that as give 100% all the time. He has no 99% mode in him - your best is to die, not literally, but burnt out, falling over the line. The limit." - Shane Legg, DeepMind cofounder
There is a great contrast at the end of the introduction from Geoff Hinton, another Nobel-winning luminary in AI (neural networks specifically), with his warning that people will be tempted to abuse AGI, but especially to build it once they know they can. It mirrors Oppenheimer's line that when you see something that is technically sweet, you go ahead and do it. Could we stop a step away?
His early career is super interesting. Obsessed with computers as well as chess as a boy, he started writing games. As a teenager he interned at a wild game startup (Bullfrog, run by the maverick Peter Molyneux). He was extremely successful, co-writing the awesome Theme Park games (I loved these) that featured very early AIs. The Bullfrog owner offered 17-year-old Demis £500k, a huge sum for a teenager. He wrote out the cheque and handed it to him to tempt him to stay and work on the next game, but he wanted to go to uni like his scientific heroes. He loved a movie telling the story of Watson & Crick at Cambridge, so that's where he went.
After excelling at uni he would do something strange for his cohort - start his own company, a game company. Here too he had early success, but then his vast ambition took him and new collaborator David Silver down the road to failure. A hugely ambitious new game would fail to materialise and the company would fold. It gave Demis his first tastes of real leadership, fundraising, and big business. All crucial later.
It was only now that he went to do his PhD. Afterwards, once again, he would turn down incredible money in the video game world as he planned to move on to his life's work, solving AI. The academic world did not seem the right fit for this; he knew he would need a lot of compute, a lot of funding, so...
DeepMind
After struggling to find like-minded collaborators during his PhD, Demis engineered some luck and introduced himself to kiwi researcher Shane Legg, a man deeper into the existing AI world. They clicked, and needed one more piece to found DeepMind - Mustafa Suleyman, now Microsoft AI CEO. He was a childhood friend of Demis's brother and has an incredible personal story: from further down the social ladder than Demis, he is presented as fearless, principled, and driven. He earned his way to Oxford only to drop out. I've read his book and never knew either this background or the manner of his DeepMind departure, both covered well here.
DeepMind was very much Demis's baby; the other cofounders did not have equal ownership shares. Demis was here in founder mode. Using his "Jedi mind tricks" and his amazing ability to storytell, he would convince and inspire researchers to join and to excel. His early experience working at Bullfrog and at his own game startup shines through - he gave people more freedom to tinker and play. Mallaby points out that there is a fine line between inspiration and control, dark and light.
Needing finance beyond his own limited funds and what was on offer in the UK, the boys headed to America.
"Demis is a true entrepreneur, he would do it for free. He will never quit. Big ambition." - Peter Thiel
Founders Fund would be the primary early backers, seeing the contrarian bet - a moonshot idea led by a brilliant founder. This was long before the AI we have today, before transformers and LLMs; it really was an expensive long shot at the time.
The pitch was a blend of neuroscience and computing, putting together neural nets and reinforcement learning, two previously polarised AI camps. DeepMind's roadmap was prescient. One aim was to ground intelligence in a world model, as without this a system could not be truly intelligent like humans. Later aspects of the neuroscience were dropped, and while world models are not yet in place, LLMs have taken off. Their projections are on track. The ~2030 AGI forecast now looks almost conservative, but it was radical 15 years ago. Shane Legg predicted it for 2028 even before DeepMind.
DeepMind was up and running, the best researchers were attracted and they worked on fundamental research and big flashy breakthroughs - often in games.
Move 37 & Protein Folding
Go is a deceptively complex board game that is hugely popular in the Far East. Unlike chess, computers were not able to defeat the best humans. Demis set his old friend and reinforcement learning specialist David Silver and team on the task. They made great progress and Demis set up a match against Lee Sedol, one of the highest-ranking players, in a televised and highly publicised event with a $1 million prize.
DeepMind's program AlphaGo would win the series 4-1. But it is move 37 of game 2 that is famous. The move was considered a mistake, something no good player would do. Only late in the game did its genius become apparent. It was creative, something outside of established play, a new way to play. 18 months later a new version, AlphaGo Zero, would beat this version 100-0. It was special as it used no human training data; it learned from playing itself and it blew past not only the best humans but the best AIs trained on human play.
High on the success, Demis pivoted many from the team onto a project he had been thinking over since his uni days, a real scientific problem: the structure of proteins.
The shape of the problem was good for AI - huge labelled datasets, clear input and output, and a blind-marked benchmark, CASP, that ran every few years. In 2018 they won the competition, a little better than anyone else. In 2020 they completely dominated, causing the surprised organisers to say protein folding for single-chain structures was essentially solved. A true breakthrough, this won Demis and project lead John Jumper the Nobel Prize in Chemistry in 2024. Demis, the son of a Greek-Cypriot father and Chinese-Singaporean mother, was knighted the same year, one of the UK's highest honours.
Google & Business
Much of the funding for this research was only possible because DeepMind was sold to Google in 2014. Outgrowing Founders Fund despite some late Elon Musk attempts, and having skilfully avoided Mark Zuckerberg, Demis decided to go with Google as he thought it was the best path to realising his overall goal. The cofounders adroitly negotiated key safety frameworks and maintained a lot of independence.
The early years brought the successes above, but tensions and reality would bite later on.
Google researchers would invent the transformer, the key technological step that powers LLMs. This is brilliantly explained by the author. They would not press home their lead here and missed the early revolution. There was more to LLMs than language; scaling them brought reasoning and understanding about the world. Demis is humble enough to know he missed this.
Google also had a version of ChatGPT a year before OpenAI, but buried it due to worries around hallucinations and the core search business. A massive miss given how things went, but it is understandable - I thought the same writing a blog post on how their search business was threatened by Perplexity and the best LLMs. Google is doing better than ever even with second-rate LLMs, so far.
Demis and Google were blindsided by Sam Altman and his rush to productise ChatGPT and his global publicity tour. Demis was furious at the starting gun of a race being fired; he felt that he too simply had to race, also publicly. This is the cold logic of technological determinism and capitalism. Google powered back in product space, by no means Demis's home field, but ever the competitor he was trying hard as part of Google. I feel you can still see this in his public comments today: he performs for the business, and is excited for the deeper science. Bard would launch already behind OpenAI. In some areas they would catch up, in narrow areas like video even surpass. But the vibes have never really shifted. Now, post Demis leaving as CEO, it is unclear if they will return to the AI frontier.
This relative success was despite Google having competing internal AI teams; Google Brain was run as a separate function outside of DeepMind for years, hampering overall progress. They would fail with Gaia, an attempt at world models, before language models powered ahead.
I was struck by the role of Sundar Pichai, Google CEO. Demis coveted independence and worked on a DeepMind spin-out, even convincing Larry & Sergey, but he was outmanoeuvred by the softly spoken and non-confrontational Sundar. Highlighting his vision with AI at the core of Google, Sundar would tease and delay and block spin-out ideas. Now we see a DeepMind VP (not CEO) reporting to Sundar, with Demis moved upstairs to Chair of DeepMind and Chief Scientist at Google. I think they wanted to keep him as long as possible and ideally never lose him, but they also need Gemini to succeed as a core part of the Google offer. Unlike, say, Waymo. Smart moves from Sundar; he's not CEO of Google for no reason.
Big Plans
There are lots of great philosophical quotes to go with those from the famed scientists and mathematicians. You will hear from the likes of Spinoza, Einstein, Kant, Gödel and Nagel, and read about their (and others') big ideas - brains in vats, simulation theory, the universe as God. Fun stuff. Taking a step back you can see Demis aiming to place himself in this tradition, the next step, and maybe one of the last we take as a species without machine intelligence.
It is in this domain he sees his future. He would like to solve intelligence, and from that disease, energy, physics, everything.
"Humanism, spiritualism, and science all go together... like Spinoza, Einstein, Da Vinci, it's art and science... it's fluid, everything is a river. Philosophy is a way of life. Self knowledge, knowledge, from sand and copper we get semiconductors... God's design... the flow is going towards finding out, understanding, I'm part of that and it's exhilarating" - Sir Demis
When asked how he spends his cash: not on fancy cars, houses or yachts, his only indulgences being some first edition books and Liverpool FC tickets (up the Reds!).
He wants to build a massive particle collider in space to help answer the big questions in physics ("what I cannot create, I do not understand" - Feynman). He wants to know if the universe is quantum or not, how we can describe the universe at the base level. He is a classical computer champion, positing that we don't need quantum explanations and that Turing machines can explain it all in 1s and 0s if powerful enough. He sees quantum mechanics as an inefficient way to run the universe and would like to cut out the weird stuff.
These ideas and goals are the great promise of artificial intelligence. But we will end on the threat.
"AGI is coming, we are not where we hoped we would be but it's coming" - David Silver
There are catastrophic and existential threats posed by AI. My p(doom) - the chance of things going this wrong in the next decade or so - has recently moved up to ~15%. I'm not alone in saying we might kill everyone, end civilisation, or lose human control over it very soon. I feel better knowing how seriously Demis takes the issue; I worry the conditions now in play raise the risk. I commend OpenAI's recently announced pause at the frontier to allow alignment work to catch up. We need a lot more work here at technical, business, and political levels to keep us on a happy path.
October 12, 2026
AI kan veilig ontwikkeld worden. Wat een tragisch boek. Gelukkig ben ik maar een pessimist.
May 15, 2026
Demis is one of the most important people in AI. Regardless of how you feel about this technology, its impact on our world is unmistakable—whether good or bad. It is, and will only continue to be, a co-author of our future. Read this book to better understand why and how.
September 29, 2026
Honestly, this is essential reading right now. More than a history of Hassabis, it is a history of the developing development of one of the most controversial and consequential technologies in human history and all the (also controversial) players involved. So much could happen in the next few months that it is hard to say how well this will or won't age. Nonetheless I loved it and learned a ton. Highly recommended.
June 19, 2026
It leans toward worship and doesn't press the ethical contradictions. It raises hard questions, then leaves you empty-handed. Little here is new if you already follow the field, and it's still too long for the story it tells.
August 16, 2026
A wonderful, deeply thought-provoking work that is both the biography of one of the most important characters on AI, and a near-complete history of the first few years of the Singularity. Highly recommended.
July 12, 2026
The narrative is at its strongest when it outlines the interpersonal politics of the competing AI labs as opposed to its high-level technical explanations of the varying models. Although I was not rapt, I guess it’s enlightening to know that 98% of the people building AI think it’s capable of overtaking humans on all tasks and possibly even destroying humanity… but at this point, what the heck am I even supposed to do with that information!?
Anyways - this is more of a 3.25, not due to a lack of quality but simply because I often felt unmotivated to pick this book back up.
Anyways - this is more of a 3.25, not due to a lack of quality but simply because I often felt unmotivated to pick this book back up.
September 24, 2026
Funny that the first office was on Russell Square, right by the street crossing where Leo Szilard had his insight about the nuclear chain reaction.
September 9, 2026
When one picks up a biographical book, one rarely expects it to be a page-turner. This book was absolutely unputdownable for me - a fascinating and, at times, frightening journey into the quest for superintelligence.
It is comprehensive, brilliantly written, and an excellent way to catch up on what on earth is going on in this AI gold rush. Sebastian Mallaby is the perfect guide, known for the extraordinary amount of research he puts into his books. If you want to understand where we are, how we got here, and where this race might be taking us, you’re in very safe hands.
It is comprehensive, brilliantly written, and an excellent way to catch up on what on earth is going on in this AI gold rush. Sebastian Mallaby is the perfect guide, known for the extraordinary amount of research he puts into his books. If you want to understand where we are, how we got here, and where this race might be taking us, you’re in very safe hands.
May 13, 2026
Hagiographic and structurally uninspiring. The way Mallaby lays out the chronology is frustrating, it mostly follows the timeline but then jumps around enough to make it confusing.
Large swaths just felt like Mallaby taking dictation from, usually, Hassabis or another source. I get that it’s a valuable resource that he had so much direct success but it seems like the methods outshined the ultimate artifact. I’m personally not a fan of Mallaby inserting his process so much talking about himself emailing and sitting down with his sources. At points it just devolves into Q&A. I thought I’d be getting more of a straight history with Hassabis as the main focus.
Main it was just the narrator of my audiobook but the end result for me was that I found it quite boring, when the subject should be riveting. Felt the same way about The Power Law, and leaves me wondering how these two books can be so much worse than More Money than God.
Large swaths just felt like Mallaby taking dictation from, usually, Hassabis or another source. I get that it’s a valuable resource that he had so much direct success but it seems like the methods outshined the ultimate artifact. I’m personally not a fan of Mallaby inserting his process so much talking about himself emailing and sitting down with his sources. At points it just devolves into Q&A. I thought I’d be getting more of a straight history with Hassabis as the main focus.
Main it was just the narrator of my audiobook but the end result for me was that I found it quite boring, when the subject should be riveting. Felt the same way about The Power Law, and leaves me wondering how these two books can be so much worse than More Money than God.
October 7, 2026
After establishing his first computer games company (Elixir Studios and game like Republic: The Revolution) and getting investor funding for it he concluded that technology is not yet ready to run the required complexity. Realizing that contemporary technology was not yet ready for his ambitions—and concluding that true artificial intelligence required a fundamental breakthrough—he shut down Elixir Studios in 2005. He returned to academia, enrolling at University College Londonfor a PhD in cognitive neuroscience. He dedicated his research to understanding how the human brain processes memory and imagination, intending to use biological cognition as the blueprint for building the AI models that would eventually define DeepMind... (true story, and there are many more that give him the "not a normal person" vibe).
This book is the other side of "Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI" that I read last year and the two complement each other well (interesting to see how both sides are seen as villains or saviours depending on viewpoint). It was very interesting insight into the history of Deepmind and Hassabis. First part focuses more on Hasabis himself but the second half already becomes extremely fast paced because instead a blue ocean is replaced with extreme competition and mixed motivations driving the participants to a death race (which we are observing still in present moment).
Key insight and reviewers thougts:
The Oppenheimer Parallel & The "Sweetness of Discovery": A prominent takeaway for many readers is how the book frames the AI race through the lens of human curiosity. Scientists push the boundaries of AI not just for profit or power, but because they cannot resist the "sweetness of discovery". This creates an "Oppenheimer problem"—once inventors know something is technically possible, it is nearly impossible to stop them from finding out where it leads, even if the resulting technology ends up controlling them.
An Outsider's Perspective: Reviewers frequently highlight that despite DeepMind being the engine room for Google, this is fundamentally not a Silicon Valley story. Demis Hassabis operates as an outsider; he takes the Valley's money but remains fiercely critical of its leaders and culture.
The "Grounding Problem" and Language Models: The book reveals why Hassabis was initially skeptical of large language models like ChatGPT. He believed AI needed to interact with the physical world to truly understand it—for example, knowing that a dropped glass will shatter requires experience, not just reading Wikipedia. However, he later realized that because the humans training these models are grounded in reality, the language they produce is inherently grounded.
The Finiteness of Human Experience: A profound realization highlighted in the text is Hassabis's conclusion that human behavior isn't as infinitely varied as we like to think. By ingesting a few trillion tokens of text, AI systems have been able to map out and understand nearly the entirety of human civilization's experiences, leading him to label language models as "unreasonably effective".
The topic of trust was covered in depth in multiple iterations
“Safety isn't about governance structures," he went on, "I mean, if you have a governance board, it probably wouldn't do the right thing when it came to the crunch. Same thing with a safety charter. You can try to negotiate one. But it's not realistic to create bright-line principles years in advance because you'll probably draw the lines in the wrong places. So discussing these things didn't really help," Hassabis continued, "It made it harder to build useful trust, because when you negotiate a trustless structure, it implies you can't trust the other person."
“Who would've thought you can actually inspire people too much?" he reflected years later. "Well, you can, because you can get to the point where you are deluding your team, and they are deluding you also. It's like, I'm making the judgement this is possible because the engineers are telling me it's possible; but they're only telling me it's possible because I've over-inspired them," Hassibis said. "So, in fact, none of us is getting real feedback.”
“if you fed an infinity of data about cells into an inductive computer, the machine might figure out a way of describing what was going on -- it would see the unseen patterns, the hidden laws, that explained cellular behavior. ”
"Human beings lack a vocabulary with which to process existential threats; they cannot think the unthinkable. They are wired to act as though life will carry on." — Sebastian Mallaby
"By ingesting a few trillion tokens, these systems have learned enough to understand nearly all of our experience... That's why I call these language models unreasonably effective." — Demis Hassabis
"The inventors believe they control their technology, but often, the technology controls them." — Sebastian Mallaby
"It's an entrancing and revealing journey, and what he finds is at once inspiring and terrifying." — Steve Coll (Reviewer)
"Once you know a thing can be done, resisting the urge to find out where it leads is very hard. Which is why I came away doubting the AI race can simply be stopped."
This book is the other side of "Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI" that I read last year and the two complement each other well (interesting to see how both sides are seen as villains or saviours depending on viewpoint). It was very interesting insight into the history of Deepmind and Hassabis. First part focuses more on Hasabis himself but the second half already becomes extremely fast paced because instead a blue ocean is replaced with extreme competition and mixed motivations driving the participants to a death race (which we are observing still in present moment).
Key insight and reviewers thougts:
The Oppenheimer Parallel & The "Sweetness of Discovery": A prominent takeaway for many readers is how the book frames the AI race through the lens of human curiosity. Scientists push the boundaries of AI not just for profit or power, but because they cannot resist the "sweetness of discovery". This creates an "Oppenheimer problem"—once inventors know something is technically possible, it is nearly impossible to stop them from finding out where it leads, even if the resulting technology ends up controlling them.
An Outsider's Perspective: Reviewers frequently highlight that despite DeepMind being the engine room for Google, this is fundamentally not a Silicon Valley story. Demis Hassabis operates as an outsider; he takes the Valley's money but remains fiercely critical of its leaders and culture.
The "Grounding Problem" and Language Models: The book reveals why Hassabis was initially skeptical of large language models like ChatGPT. He believed AI needed to interact with the physical world to truly understand it—for example, knowing that a dropped glass will shatter requires experience, not just reading Wikipedia. However, he later realized that because the humans training these models are grounded in reality, the language they produce is inherently grounded.
The Finiteness of Human Experience: A profound realization highlighted in the text is Hassabis's conclusion that human behavior isn't as infinitely varied as we like to think. By ingesting a few trillion tokens of text, AI systems have been able to map out and understand nearly the entirety of human civilization's experiences, leading him to label language models as "unreasonably effective".
The topic of trust was covered in depth in multiple iterations
“Safety isn't about governance structures," he went on, "I mean, if you have a governance board, it probably wouldn't do the right thing when it came to the crunch. Same thing with a safety charter. You can try to negotiate one. But it's not realistic to create bright-line principles years in advance because you'll probably draw the lines in the wrong places. So discussing these things didn't really help," Hassabis continued, "It made it harder to build useful trust, because when you negotiate a trustless structure, it implies you can't trust the other person."
“Who would've thought you can actually inspire people too much?" he reflected years later. "Well, you can, because you can get to the point where you are deluding your team, and they are deluding you also. It's like, I'm making the judgement this is possible because the engineers are telling me it's possible; but they're only telling me it's possible because I've over-inspired them," Hassibis said. "So, in fact, none of us is getting real feedback.”
“if you fed an infinity of data about cells into an inductive computer, the machine might figure out a way of describing what was going on -- it would see the unseen patterns, the hidden laws, that explained cellular behavior. ”
"Human beings lack a vocabulary with which to process existential threats; they cannot think the unthinkable. They are wired to act as though life will carry on." — Sebastian Mallaby
"By ingesting a few trillion tokens, these systems have learned enough to understand nearly all of our experience... That's why I call these language models unreasonably effective." — Demis Hassabis
"The inventors believe they control their technology, but often, the technology controls them." — Sebastian Mallaby
"It's an entrancing and revealing journey, and what he finds is at once inspiring and terrifying." — Steve Coll (Reviewer)
"Once you know a thing can be done, resisting the urge to find out where it leads is very hard. Which is why I came away doubting the AI race can simply be stopped."
Displaying 1 - 30 of 404 reviews