Google DeepMind CEO Demis Hassabis’s AI journey had an unexpected start: his early mastery of chess.
Years before Hassabis would receive the 2024 Nobel Prize in Chemistry for creating an AI program that predicted protein structures, he was a child chess champion who started playing the game at the age of four. By age 13, he was a chess master competing against adults in international competitions.
In a lecture earlier this month at the University of Cambridge, Hassabis, now 48, explained that chess got him “thinking about thinking itself” or exploring the mental processes behind complex thoughts.
“How does our mind come up with these plans, these ideas?” Hassabis asked. “Perhaps more fascinating to me than the games I was playing was the actual mental process behind it.”
Hassabis’ first exposure to programming as a child was through an electronic chess computer, a physical board capable of playing chess against a human player. Though Hassabis was meant to test out different chess strategies on the computer, he was more interested in how the computer worked and how someone had programmed it to play chess.
“I remember being fascinated by the fact that someone had programmed this lump of inanimate plastic to play chess really well against you,” Hassabis said. “I was really fascinated by how that was done and how someone could program something like that.”
In his early teens, Hassabis began trying to build AI programs himself on an early home computer, the Amiga 500. From then on, he “was hooked” on AI and decided to spend his entire career making advances in the field.
Demis Hassabis. Photographer: Stefan Wermuth/Bloomberg via Getty Images
Hassabis co-founded AI company DeepMind in 2010, and it was acquired by Google in 2014 for more than $500 million. He later invented AlphaZero in 2017, an AI algorithm that needed only the rules of chess and four hours of playing against itself to become the strongest chess player ever, beating human chess masters.
Hassabis was awarded the Nobel Prize in Chemistry in 2024 with fellow laureate and DeepMind director John Jumper for creating an AI model, AlphaFold2, that accurately predicted the complex structures of almost all 200 million proteins, each within minutes. The AlphaFold Protein Structure Database, which makes these protein structures freely available, has reached over two million users in 190 countries, helping pave the way for advanced research in areas like Parkinson’s treatments and antibiotic resistance.
Hassabis pointed out in his Cambridge lecture that it takes an average of 10 years, plus billions of dollars, to create a new drug. According to the London School of Economics and Political Science, the average cost of developing a drug ranges from $314 million to $2.8 billion.
In the lecture, Hassabis touched on the possibility of developing drugs more quickly and cheaply using AI, “from potentially years down to minutes and seconds.”
Hassabis told DeepMind employees in London earlier this month that he thinks artificial intelligence will become more intelligent than human beings within the next decade.