Jurgis Posted February 14, 2017 Share Posted February 14, 2017 Since we have a bunch of poker players and a bunch of Canadians on CoBF, this might be interesting: AI wins poker tournament: https://arstechnica.com/gaming/2017/01/robot-knows-when-to-hold-em-wins-huge-in-poker-tournament/ Most articles don't mention this, but currently poker AI can handle only one-vs-one play. It cannot handle one-vs-many play yet. U Alberta poker research ( http://poker.cs.ualberta.ca/ https://arxiv.org/pdf/1701.01724v2.pdf ) promises to run on laptop (vs. CMU supercomputer above) and still win. Disclaimer: I am not a poker player and not an expert in the research above. Link to comment Share on other sites More sharing options...
rtvinvest Posted February 15, 2017 Share Posted February 15, 2017 This is actually highly impressive. The bot (computer program) managed to beat 4 of the best human 1-on-1 poker players on the world by quite a big margin over a large sample size of hands played. During the last version of this match up the human team easily beat the bot, but they mentioned the program had improved significantly this time around. As the 'holy grail' in poker is to approach a game theory optimal (GTO) strategy raw computing power which learns by playing itself many times over can help speed up this process. (fun fact: some of the best players have actually hired programmers / servers in order to build custom software to help them solve the game) Link to comment Share on other sites More sharing options...
wachtwoord Posted February 15, 2017 Share Posted February 15, 2017 No time to read right now but this is likely limit holdem. A far easier game for AI to beat than the more popular no limit holdem. Link to comment Share on other sites More sharing options...
rtvinvest Posted February 15, 2017 Share Posted February 15, 2017 It's not. Its heads-up no limit holdem vs 4 (likely to be top 10 globally) human players. Link to comment Share on other sites More sharing options...
wachtwoord Posted February 15, 2017 Share Posted February 15, 2017 Wow this is extremely impressive (I just looked at the paper). Especially considering it was 200bb too. Extremely impressive. this is a very very hard problem to solve. Thanks for sharing this. Link to comment Share on other sites More sharing options...
CorpRaider Posted February 15, 2017 Share Posted February 15, 2017 Awesome. What is the ETA on a not horrendous spelling auto-correct? Link to comment Share on other sites More sharing options...
rkbabang Posted February 15, 2017 Share Posted February 15, 2017 Awesome. What is the ETA on a not horrendous spelling auto-correct? +1. You'd think that if AI can play poker that it could use context clues to determine what wood we meant to type. Link to comment Share on other sites More sharing options...
Jurgis Posted February 15, 2017 Author Share Posted February 15, 2017 Fnuny you aksed: http://export.arxiv.org/pdf/1608.02214 In general the spelling checking/correction state of the art is likely way better than what you get, especially in a browser editor. However, there's likely hundreds of "spell checker" codebases in the world and probably none (well maybe 1-2) use anything close to latest technology. If I had to guess, some mobile typing apps should have close to latest technology, but I can't say much more. Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now