Poker? I don't even know 'er!
Anyway, no existing artificial intelligence could understand that joke. Likewise, I think Scott Chaffin at The Fat Guy is right in thinking that no existing AI could beat a human poker master. However, I think he's hasty in dismissing speculation that people are writing bots to play online poker and are raking in the dough from unsuspecting internet players.
Are poker ‘bots’ raking online pots?: after "these online sites are RIGGED!" and "I'm calling tech support about collusion!", bots are probably the biggest source of idle chit-chat and/or creeping Nixon-like paranoia on the interweb. As this article explains, it's highly unlikely that you're sliding pixelated chips to software programs. There's talk of one up in that Canada, called Vex something-or-other, that is rated a "master" at 2-handed games. Well, a) big whoop: only a dumbass plays two-handed unless they're squaring off against someone they know (we're talking regular ring games here, not tournies), and b) "master"? What the heck is that? By my reckoning, the only masters of this game are walking around Vegas with rolls of $1000 bills that would choke a horse and getting on the teevee on a semi-regular basis (cf., Doyle Brunson, TJ Cloutier, Daniel Negreanu, etc., etc., etc.) All that is to say I don't spend a lot of time worrying about robots. If someone wants to give their judgement over to a piece of software when they're playing for real money...well, let's just say I've written software, and poker ain't completely logical.I've written software too -- and am getting a Ph.D. in artificial intelligence -- and I'm positive that it wouldn't be hard to write an AI that could beat average to good poker players in an online format. "Vex something-or-other" probably refers to Vexbot, a component of Poki's Artificial Intelligence, a project led by Darse Billings from the University of Alberta.
According to the Poki FAQ:
Q: How good is Poki?Sounds reasonable to me. If Poki ever can beat human poker masters then that means that bluffing and "tells" are unnecessary to winning the game. Poki probably does "bluff", but only based on statistics, not intuition, and that seems like a very different thing to me. Imagine how effective an AI could be that could read human facial expressions in addition to crunching numbers?A: The older version of Poki that plays in full 10-player games is better than a typical low-limit casino player, and wins consistently against average opponents; but it is not as good as most expert players. The newer programs being developed for the 2-player game are quite a bit better, and we believe they will eventually surpass all human players, perhaps within a few years, or less.
Heh. Just to be clear, I want to clarify my comments on poker-bots. I'm talking about commercially available (or close) Windows programs that might be squaring off against me and my pals in online poker rooms. I don't beleive it's there yet.
I have no doubt that two-handed bots will eventually be good enough to beat the top dogs. But top dogs don't play poker two-handed against computers. They play against schmoes like me so they can take our money.
I also believe there are bots that can play Group 1 hands and win money at the online tables. My wife can do that, if she'd bother to do so. We even call it Robot Poker when we're doing that. What I don't believe they can do yet is play the hands that make REAL money -- the marginal ones that require lots of post-flop maneuvering based on number of hands being played, bet sizes, position, session patterns, and lifetime patterns. That's a deeply complex 'thing'.
With any luck at all, I'll be long gone from the online tables when the really good bots get here ;) I'm sure that there will be a new breed of player who will be thrilled to face off against bots for the competitive challenge. Not me, though.
Anyhoo -- fun stuff to think about.
SC: I gladly yield to your superior poker experience as to the complexity of hands that make "REAL money". I'm sure the problem is quite difficult, and I didn't mean to imply that Mr. Billings' project was exhaustive or trivial. Still, I think that the advantage a computer has with pure number-crunching outweighs the intuition it lacks. Calculating pot odds, draws, outs, and learning the characteristics of different types of players can all be done incredibly accurately by a computer, and it won't make mistakes. The details are less tangible than in, say, chess, but the move space isn't any larger (excluding the differences between $10 raises and $11 raises, for instance).
I personally wouldn't play any game online for money because I don't entirely trust the casino operators, not to mention bots and colluding players. Still, if you make net a profit from it then I can only assume you're onto something.
I've heard rumors floating around the casinos about a major poker site which used bots in a high limit poker game. The intent was not to win money but to keep the game going long enough to establish a solid player base to support a $100-200 omaha 8 or better table. However the players figured it out and exploited the systematic nature of the a.i. for over $100,000. Or so the story goes.