table of contents
Table of Contents Ask a question to the chatbot
overview
Born: July 27, 1976, London, England (48 years old) (more)
Demis Hassabis (born 27 July 1976 in London, England) is a British computer scientist who won the 2024 Nobel Prize in Chemistry for his work using artificial intelligence (AI) to predict the structure of proteins. was awarded. He shared half of his prize money with a fellow American computer scientist. John M. Jumper, and the other half of the prize went to an American biochemist. David Baker.
Early life and career
Hassabis spent his childhood in north London. Born to a Greek-Cypriot father and a Chinese-Singaporean mother, Hassabis developed a keen interest in chess and computer programming. A chess player since the age of four, Hassabis earned the rank of Chess Master at the age of thirteen. His love for computer games inspired him to code his first game, a version of Othello, and he taught himself computer programming with books.
Hassabis continued writing code for the game company Bullfrog Productions, and by the age of 17 he was the lead developer on the AI-generated video game Theme Park (1994). During this period he attended grammar school at Queen Elizabeth School and secondary school at Christ College, Finchley. At Finchley College, I completed my exams two years early at the age of 16.
In the 1990s, Hassabis studied computer science at Cambridge University and was captain of the university’s chess team. After graduating from Cambridge in 1997, Hassabis became the lead AI programmer at Lionhead Studios, where he helped develop the game Black & White (2001). He left Lionhead Studios in 1998 and founded Elixir Studios, which has produced award-winning AI-based video games including Republic: The Revolution (2003) and Evil Genius (2004).
After selling his stake in Elixir Studios in 2005, Hassabis took his passion for AI algorithms in another direction. He pursued a Ph.D. I completed my PhD in Cognitive Neuroscience at University College London (UCL) primarily to understand how the human brain manages imagination and memory. After graduating from UCL in 2009, Hassabis did postdoctoral work at Harvard University and the Massachusetts Institute of Technology (MIT), and became a Henry Wellcome Fellow at UCL.
deep mind
Hassabis founded the AI-based startup DeepMind in 2011 with New Zealand computer scientist Shane Legg and British entrepreneur Mustafa Suleiman. Over the next four years, he developed a series of AI models and used deep learning to train them and enable playback and playback. Master video games without teaching AI the rules. Hassabis and his colleagues sold DeepMind to Google in 2014, and Hassabis remained CEO.
DeepMind AI became the platform for AlphaGo, which defeated top Go player Lee Sedol 4-1 in 2016. AlphaGo’s neural network learned how to play Go from human players and by playing the game itself. AlphaGo was overtaken by AlphaGo Zero, which started with just the rules of Go. AlphaGo Zero shut out AlphaGo 100 games with 0 losses. Alpha Zero, a more general neural network, quickly learned to play chess and shogi using the same technique.
AI solves protein folding
Despite the media attention generated by DeepMind’s gameplay AI victories, Hassabis is actually applying what he learned from those AI successes to scientific problems, specifically the problem of protein folding. I was interested in that. Proteins are large molecules that are directly involved in the chemical processes essential to life and are made up of 20 amino acids that can be combined in a variety of ways. A protein’s function is determined by its highly complex three-dimensional structure, based on how strings of amino acids are folded.
How a protein folds is determined by its amino acid sequence. But even a small protein with just 100 amino acids can have 1,047 possible three-dimensional structures. Predicting protein structure from amino acid sequences has become an important problem in molecular biology.
In 1994, biologists John Molt and Krzysztof Fidelis founded the Critical Assessment of Protein Structure Prediction (CASP), a challenge to test methods for predicting the structure of proteins. Every two years, contestants were given the amino acid sequence of a protein whose structure had been determined but had not been published, and were asked to predict the protein’s structure.
Progress was slow. By the mid-2010s, the best models in the CASP challenge had an accuracy of about 40%. DeepMind joined the protein structure program AlphaFold at CASP13 in 2018, achieving an impressive accuracy of approximately 60%, far exceeding its competitors. But further improvements were difficult to make until Jumper joined DeepMind and used his protein simulation experience to help develop AlphaFold2.
AlphaFold2 was trained on a database of amino acid sequences and protein structures and used a neural network called a transformer to find possible protein structures. In CASP14 in 2020, AlphaFold2 reached 90% accuracy, which is comparable to experimental results. The problem of finding the structure of a protein from an amino acid sequence has been solved.
Hassabis, Jumper, and their collaborators used AlphaFold2 to calculate the structures of nearly all of the more than 50,000 human proteins in 2021. They went further and calculated the structures of nearly all of the 200 million known proteins, originating from ca. Millions of different species, or in Hassabis’ words, “an entire universe of proteins.”
Hassabis has received numerous honors, including the Royal Society’s Mourad Prize (2014), the Pontifical Academy of Sciences’ Pius XI Medal (2020), and the Wiley Prize in Biomedical Sciences (2022). In 2023, he received the Life Sciences Breakthrough Award, the Canada Gardner International Award, and the Albert Lasker Basic Medical Research Award.
Eric Gregersen John P. Lafferty
