This week Deep Mind – the Google team behind the “Alpha Zero” artificial intelligence project – tried out the game of chess on the system.
Alpha Zero spent four hours learning to play chess from scratch – essentially playing against itself – before being pitted against the result of perhaps forty years of chess software research and development (an incredibly strong chess engine called Stockfish). Alpha Zero destroyed Stockfish again, and again.
The interesting thing? It didn’t play like a computer at all – it played like a Grandmaster from the “romantic” era of chess, perhaps a hundred years ago – taking enormous risks, and throwing pieces away in pursuit of victory.
Food for thought.