✳flâneur — a map of the web's best reading

Playing chess with large language models

nicholas.carlini.com · 2,385 words · saved by 1 readers

Computers have been better than humans at chess for at least the last 25 years. And for the past five years, deep learning models have been better than the best humans. But until this week, in order to be good at chess, a machine learning model had to be explicitly designed to play games: it had to be told explicitly that there was an 8x8 board, that there were different pieces, how each of them moved, and what the goal of the game was. Then it had to be trained with reinforcement learning agaist itself. And then it would win. This all changed on Monday, when OpenAI released GPT-3.5-turbo-instruct, an instruction-tuned [a] language model that was designed to just write English text, but that people on the internet quickly discovered can play chess at, roughly, the level of skilled human players. (How skilled? I don't know yet. But when I do I'll update this!) You should be very surprised by this. Language models ... model language. They're not designed to play chess. They don't even kn

Playing chess with large language models Main Papers Talks Code Writing Writing Playing chess with large language models by Nicholas Carlini 2023-09-22 Computers have been better than humans at chess for at least the last 25 years. And for the past five years, deep learning models have been better than the best humans. But until this week, in order to be good at chess, a machine learning model had to be explicitly designed to play games: it had to be told explicitly that there was an 8x8 board, that there were different pieces, how each of them moved, and what the goal of the game was. Then it

Explore this link on the map →

saved by

related reading