Once upon a time humans gave a machine all the known recorded moves of chess grand masters, and the machine used the information to beat the world champion chess player Gary Kasparov.
Adam Kucharski built a machine to play poker against. Poker isn’t like chess. There is no compendium of moves. Kucharski set up the machine up to learn the game. He was successful but he didn’t have much fun. The machine beat him every time.
The machine learns from its mistakes. When Alan Turing thought about machines that could think, the analogy was a child’s mind. Adults look forward. Children look back and ask themselves ‘what would I do differently if I had to do that again’.
Poker is complicated, and the machine, like the human, abstracts, simplifies. But humans tend to go too far with the abstraction. Machines figure out what it takes to win, and in games of poker that means being a bit weird sometimes.
Machines don’t do emotion. Humans do. And emotion isn’t great for games like poker.
As a species, we are never again going to beat machines at games like chess and poker. But it seems reasonable to suppose that we may learn from the way machines think. Looking back and learning from our mistakes. Not over simplifying, becoming a little unpredictable. And maybe a little less emotional?
Adam Kucharski, The Perfect Bet