My Review of Sebastian Mallaby's "The Infinity Machine: Demis Hassabis, Deep Mind......."

 The Prophet of AI

 

Sebastian Mallaby, Council on Foreign Relations fellow, tells us the story of artificial general intelligence through the life of Demis Hassabis, the founder of Google’s Deep Mind. The story is not limited to Hassabis as Mallaby discusses his rivalry with Sam Altman’s OpenAI and Mark Zuckerberg's Meta. Along the way we meet venture capitalist Peter Thiel and Elon Musk. Mallaby also goes into great detail as to how AI works and how AI agents can hallucinate and go rogue. Thus, in a way the book is primer on AI.

 

Hassabis is the son of immigrants to North London, with a Chinese Singaporean Mom and a Greek Cypriot father. As a child he was chess prodigy and was a champion by age 13. From chess Hassabis moves on to gaming, first a player and then as a successful designer. He gets degrees from Cambridge in physics and neuroscience which gives him an excellent background to pursue his interest in AI.

 

He forms Deep Mind in 2010 with the support of Peter Thiel and other venture capitalists. There he and his colleagues do pioneering work in deep learning and reinforced learning. In 2014 Google bought out Deep Mind for $650 million, making Hassabis worth $134 million. His interest is not in money but rather pursuing the science of artificial general intelligence. Their models became experts at the highly complicated game of Go, so much so that they defeat the world champion in a globally televised event in Korea.

 

The reinforced learning technique used for GO was then used to detect the shapes of proteins using their Alpha Fold model. This innovation is hugely important in the drug discovery process, so much so that Hassabis along with his colleague John Jumper win the Nobel Prize for chemistry in 2024.

 

One of Hassabis’ colleagues at Deep Mind was Mustapha Sulayman, another immigrant son from North London. Sulayman was very concerned about AI safety, so much so that Google delayed releasing its early AI models to the public. After an intital misstep, Google's Gemini has become a world beater. However, when Open AI released ChatGPT, the competitive juices began flowing and the safety issues were put on the back burner. Ultimately Sulayman ended up heading Microsoft’s AI efforts. In a strange quirk of history, we now two immigrant children from North London running the AI efforts of two of the major players in the field.

 

Mallaby hardly mentions Dario Amodei, the head of Anthropic, which is fast becoming an industry leader. Amodei was at OpenAI and then left to form his own company. Anthropic is the developer of Claude which is fast becoming the AI model of choice in the enterprise. My guess is that Mallaby did not have the time to interview or understood the significance of Amodei in the AI race. That aside Mallaby has written an important book on what is becoming a revolution in how we approach the world around us.