(A short blog by Alain Dekker, October 10, 2024)

We would like to congratulate Demis Hassabis for winning the 2024 Nobel prize in Chemistry! Earlier, the 2024 Nobel prize for Physics was awarded for work on artificial neural networks (ANNs) to John Hopfield and Geoffrey Hinton [1]. And in a nod to one of the many applications of ANNs, the 2024 Nobel prize for Chemistry has been awarded for work on protein structure prediction to David Baker, John Jumper and the Mind Sports Olympiads’ very own Demis Hassabis [2].

Proteins are large biomolecules composed of long chains of amino acids. Determining their 3D structure experimentally is both complex and time-consuming, so the promise of accelerating the discovery process using computers has excited researchers for years. But until recently, this has remained largely unfulfilled. The breakthrough came in 2020 with AlphaFold [3][4], developed by Google Deepmind. AlphaFold works by training complex ANN models on large amounts of known 3D structure data of proteins. On recent benchmark tests, AlphaFold was successfully able to predict the structure of most proteins based solely on the input of their 2D sequence of amino acids.

Most game players will be more familiar with AlphaFold’s predecessor applications, AlphaGo and AlphaZero. In 2016, AlphaGo made history by defeating the best Go player in the world, Lee Seedol 4-1 [5][6]. And in 2017, AlphaZero was able to beat Stockfish, the strongest “brute force” chess program at that time [7]. Since then, LeelaZero (based on AlphaZero) and Stockfish NN, have incorporated ANN learning to reach even stronger levels. [8][9]

Demis co-founded Google Deepmind in 2010, shortly after completing a PhD in neuroscience at UCL. His neuroscience research led to influential publications in Nature and provided inspiration for his work on ANNs at Deepmind. Before that, Demis worked on computer games development at Bullfrog Productions, Lionhead Studios and then his own company, Elixir Studios. [10]

Running through his entire career from computer games development, to neuroscience, AlphaGo and AlphaFold, and all the way to the Nobel prize itself, has been Demis’ deep passion for games. Demis started his boardgames career playing Chess and quickly reached master level. By the time Demis switched focus to computer games, he was one of the highest ranked 13-year old chessplayers in the world, 2nd only to Judit Polgar. In 1997, the Mind Sports Olympiad was inaugurated and Demis quickly proved exceptional at learning new games, winning the Pentamind [11] a record five times from 1998 to 2003. Side note: Demis’ record was beaten this year when Andres Kuusk won his 6th Pentamind.

Demis has always been approachable and eager to share his expertise. When I first met him in 2003, Demis taught me Entropy, a game invented by British Chemical Engineer Eric Solomon and having the distinction of being the only game (in exactly the same form), that has been at every MSO since the beginning. Demis has won the Entropy World Championship five times from 2000 to 2006, a record beaten recently by David Jameson.

Many present-day applications shaking the world of business and science are based on ANNs. These applications include AlphaFold, ChatGPT [12] and driverless cars. The list keeps growing. This work was started by Hopfield, Hinton (and others), and developed into practical, working technology by Demis (and others). The inspiration for all of this work comes from our innate human curiosity to ‘play’ with and to understand nature.

Games are fun, sociable and stimulating. As we reflect on the power, promise and risks of applications based on ANNs, we can also observe that games are serious stuff! And who knows…perhaps at next years’ Mind Sports Olympiad, you will have the pleasure of meeting Demis across the Chess or Entropy board!

### References (edit or omit as required) ###
[1] https://www.nobelprize.org/prizes/physics/
[2] https://www.nobelprize.org/prizes/chemistry/
[3] https://en.wikipedia.org/wiki/AlphaFold
[4] https://deepmind.google/technologies/alphafold/
[5] https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol
[6] https://www.theguardian.com/technology/2016/mar/15/googles-alphago-seals-4-1-victory-over-grandmaster-lee-sedol
[7] https://www.chess.com/news/view/google-s-alphazero-destroys-stockfish-in-100-game-match
[8] https://lczero.org/
[9] https://stockfishchess.org/blog/2020/introducing-nnue-evaluation/
[10] https://en.wikipedia.org/wiki/Demis_Hassabis
[11] https://en.wikipedia.org/wiki/Mind_Sports_Olympiad
[12] https://openai.com/index/chatgpt/