![]() The bots came out on top over time, despite some ups and downs along the way. In one version of the experiment, five bots played one human. The bot played 10,000 hands of poker against more than a dozen elite professional players, in groups of five at a time, over the course of 12 days. This process enabled Pluribus to hone its algorithm - its "blueprint strategy" - for the next phase, competition against human beings. If a different move would have improved the odds of winning, the bot would decide to do that move more often. It examined what the outcomes might have been had it played differently. It did this by playing hands against copies of itself. First, Pluribus had to get good at poker. Looking ahead to the end of the game "would take longer than the life of the universe," Sandholm said. Sandholm described Pluribus as a "depth-limited look-ahead algorithm." Pluribus, in deciding what to do (for example, call a bet, raise a bet or fold), calculates the odds of winning the hand, but it makes calculations only a few steps ahead rather than all the way to the end of the game, which would require implausible amounts of computation. He has started two private companies to commercialize the software, he said.Īn earlier iteration of the software program, called Libratus, had demonstrated that it could win at two-player poker, but Pluribus works in a multiplayer poker game, a far more complicated situation. Tuomas Sandholm, Brown's adviser at Carnegie Mellon and the co-author of the new paper, said the development of Pluribus comes after 16 years of research and incremental improvements in the software. The key challenge, he said, is, "How do you get AI to cope with hidden information in a complex, multi-agent environment?" "The techniques underlying it are very general and I think are going to be applied to a wide variety of settings," said lead author Noam Brown, who works at Facebook Research and is a graduate student at Carnegie Mellon University, where he began this research. Moreover, unlike the "deep learning" AI programs that became unbeatable at chess and Go, Pluribus does not use massive amounts of data and computation. It could even be used in political campaigns - helping candidates decide where to allocate resources in a contest with multiple opponents, each with a secret strategy. ![]() This technology, the inventors say, could be applied to self-driving cars, auctions, contract negotiations and decisions about product development. ![]() This milestone in artificial intelligence has implications beyond poker, or anything happening on the gaming tables in Las Vegas. It often bets huge sums early in a hand - reminiscent of the disruptive tactics of "Jeopardy!" champion and professional sports bettor James Holzhauer. But it also has some startling tendencies, including a bewildering unpredictability in its betting habits. Pluribus employs a strategy that in some respects affirms the best tactical instincts of the game's top players.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |