Libratus
In plain wordsA poker-playing AI from Carnegie Mellon University that in 2017 decisively beat four top professionals at two-player no-limit Texas hold'em. Its successor, Pluribus, beat professionals at six-player poker in 2019.
A program for heads-up (two-player) no-limit Texas hold'em poker developed by Noam Brown and Tuomas Sandholm at Carnegie Mellon University, which in January 2017 beat four leading professionals over 120,000 hands at the Brains vs. Artificial Intelligence event in Pittsburgh by a statistically significant margin. Poker is a game of imperfect information, in which hidden cards and bluffing defeat the search methods used for chess and Go. Libratus rested on three components: a precomputed 'blueprint' strategy approximating a Nash equilibriumNash equilibrium, n.: A stable standoff in a game: each player's choice is already their best answer to what the others are doing, so nobody gains by changing course alone — though the result need not be good for anyone. of a simplified version of the game, found by counterfactual regret minimization; fresh solving of each subgame as play reached it; and, between sessions, a self-improvement step that shored up the spots the professionals had been probing. Its successor, Pluribus (2019, developed with Facebook AI Research), beat elite professionals in six-player games, the first superhuman result in multiplayer poker, using far less computation.
Libratus bluffed in spots that human professionals considered reckless, and it was right.
Origin
The name is Latin for 'balanced'. Libratus was described in Science (online December 2017) in 'Superhuman AI for heads-up no-limit poker: Libratus beats top professionals', and Pluribus in Science in August 2019.
Usage note
Brown later joined OpenAIOpenAI, prop. n.: The American company that makes ChatGPT and the GPT models. It began in 2015 as a nonprofit research lab and is now a for-profit company overseen by a nonprofit foundation., where he worked on the OpenAI o1OpenAI o1, prop. n.: OpenAI's first reasoning model, from 2024. Before replying it writes out a long, hidden series of steps, trying ideas and checking them, which lifted its scores on hard math, science and coding problems far above earlier chatbots'. reasoning models, citing poker's lesson that extra computation at decision time can substitute for far larger models.
Related
- Nash equilibriumNash equilibrium, n.: A stable standoff in a game: each player's choice is already their best answer to what the others are doing, so nobody gains by changing course alone — though the result need not be good for anyone.
- game theorygame theory, n.: The mathematics of strategic decisions, where each player's best move depends on what the others do. It is used to study competition and cooperation among people, firms, nations and AI agents.
- CiceroCicero, prop. n.: An AI from Meta that in 2022 played the strategy board game Diplomacy against people online, negotiating and forming alliances in ordinary written language, and scored better than most of its human opponents.
- AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it.
- test-time computetest-time compute, n.: The computing effort an AI model spends while answering a question, as opposed to while being trained. Letting a model think longer or try many answers can make it markedly smarter.
Sources
- Brown, N. & Sandholm, T. (2018). Superhuman AI for Heads-Up No-Limit Poker: Libratus Beats Top Professionals. Science 359(6374), 418–424.
- Brown, N. & Sandholm, T. (2019). Superhuman AI for Multiplayer Poker. Science 365(6456), 885–890.