THE LINUX GAMBIT

Netflix’s hit show The Queen’s Gambit has opened more people’s eyes to the fascinating world of chess. The characters in that show may be fictional, but chess aficionados will know only too well that the ancient game has seen more than its share of colourful personalities. Ever since a machine beat Garry Kasparov in 1992, chess engines have garnered significant interest. Today, the open source Stockfish engine is considered one of the best in the world, and it regularly beats grand masters.

So it was a little shocking when Deepmind announced its neural-network based AlphaZero engine – which had trained itself (albeit on terribly expensive custom hardware) in a mere four hours – had decidedly trounced Stockfish. Stockfish uses complex heuristics created by chess dons together with brute force (on modern hardware it has no trouble analysing some 20 moves ahead), whereas, pre-training, AlphaZero ‘knew’ only the basic rules of chess and that its goal was to win.

A few of these games were released to the public, and they showcasedidn’t care about losing material so long as there was adequate compensation (which wasn’t at all obvious even to a keen player).