Yes, But Can The Computer Keep A Poker Face?

from the deep-stack dept

Computers have been beating humans at chess for a long time now, and just recently a group of scientists announced that they had developed an invincible checkers computer. But getting a computer to excel at poker has been something of a challenge. Whereas in chess and checkers, all of the necessary information is available to the computer for it to compute, in poker the players are dealing with imperfect information. Thus, good poker players often rely on feel and intuition, which are weak spots for machines. But computers are getting better, and starting today, a new poker-playing computer will square off against two poker pros in a contest to determine whether this bastion of human superiority is bound to fall. The program's developers have come up with a clever method to minimize the role of luck, as the computer will play two separate games simultaneously against the pros. The exact same cards will be dealt in each game, but in one game, the human will receive the cards that the computer got in the other game, and vice versa. In the end, the winner will be determined by combining the humans' chips and comparing them to what the computer has. Obviously, one contest won't be enough to give a definitive answer on this question, but if the computer does well, it will indicate that certain traits, like intuition, can actually be programmed to some extent.


Reader Comments (rss)

(Flattened / Threaded)

  •  
    identicon
    MiniDevil, Jul 23rd, 2007 @ 7:12pm

    Sounds fun.

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    insomniac, Jul 23rd, 2007 @ 8:32pm

    "Sound fun"
    Not really. What's the point of playing if you can't win. I still can't beet the computer at checkers or chess in Ubuntu. I've actually given up even on easy the computer kicks my ass. So what we found a way to give the computer an advantage. It already has an advantage it can process way more calculations per minute than I can.

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    Mark, Jul 23rd, 2007 @ 8:33pm

    I got 5 on it

    but... Only 2 games? How can you accurately judge with only 2 games? I'd like to see 2 years worth of games to formulate an educated guess.

    M

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    Lutomes, Jul 23rd, 2007 @ 10:40pm

    Chip count

    As long as the chip count gets reset at the start of each hand then I think its ok to start with. (obviously a limited sample) But if you have to use the winnings of your previous hands to bet your new ones (like normal poker) then it really invalidates the test. Otherwise it becomes too risky for either side to go 'all in'.

    On the other hand maybe I should read the article, because that would most likely answer my question...

     

    reply to this | link to this | view in chronology ]

    •  
      identicon
      Simon, Jul 24th, 2007 @ 7:42pm

      Re: Chip count

      But if you reset the chip count, it wouldn't be the same as real poker, where players have to balance the risk of losing many of their chips against the liklihood of winning; and where players may be willing to pay for information (e.g. call a bet on the river when they are sure they are beaten, in order to see what cards their opponent had, and how they played the hand).

       

      reply to this | link to this | view in chronology ]

  •  
    identicon
    Shohat, Jul 23rd, 2007 @ 11:19pm

    Wrong conclusion

    It will mean that you can program can make an educated guess.

     

    reply to this | link to this | view in chronology ]

  •  
    icon
    Zubin (profile), Jul 24th, 2007 @ 12:10am

    Not intuition, not a guess.

    Poker Theory says that with calculation one will be able to develop a mixed strategy that will be able to break even against any opponent in the long run.

    If the humans are unassisted, all it will take is raw computational power and sufficiently random number generation, and the computer can overpower any opponent.

    They may lose a hand from time to time, but as any poker player will tell you, everybody has their bad beats.

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    Stephen, Jul 24th, 2007 @ 6:31am

    poker

    this is like what they do in bridge tournaments. two people on a team will play east-west in one room while two more will play north-south in another. the former will play the same cards as the latter's opponents, and vice versa.

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    dorpus, Jul 24th, 2007 @ 6:42am

    Why would it be hard?

    Computers can calculate the precise conditional probabilities far better than humans can. The search tree is relatively small, regardless of the version of poker played -- at most, there are just 5 lookahead levels. It's not even big enough to warrant a Bayesian search strategy.

     

    reply to this | link to this | view in chronology ]

    •  
      identicon
      Simon, Jul 24th, 2007 @ 7:41pm

      Re: Why would it be hard?

      Conditional probabilities are not hard to calculate in poker. At least in Hold 'Em, they are very simple - if I'm drawing to a flush on the river, I have a one in five chance of making it (it's not exactly 1 in 5, but it's close enough to make no difference). If I know I have to hit a flush to win, and I have to pay more than 1/5th of the pot to call, then I should fold.

      Knowing the maths is the most basic part of poker; to be a halfway decent player you need to be able to read your opponent, understand what his/her betting patterns mean, work out when s/he's bluffing, etc. Can a computer do that? Maybe, but it's a lot harder than working out the odds.

       

      reply to this | link to this | view in chronology ]

  •  
    identicon
    jecy, Jul 24th, 2007 @ 6:55am

    I just confused how the computer to win human's mind?? As this system was designed on computer by human. I would like to play chess in computer sometime, but alway lose, that i don't have any confident to play it.

    www.acb2b.com, the free trading platform for business, joint it, you will find your potential buyer & seller.Our service include agents/distributors recruiting services, online/offline marketing campaigns, Chinese Business Credential Investigation, Chinese market research, trade show services and other services related to doing business in China.

     

    reply to this | link to this | view in chronology ]

  •  
    identicon
    Anonymous Coward, Jul 24th, 2007 @ 7:00am

    I fail to comprehend why people feel the need to create computers that can beat human beings at various games. Just what exactly are you trying to accomplish, and how does it benefit mankind? For cryin' out loud, use those super computers to find a cure for cancer, or for some other productive task. A game is no longer a game when you to rip it apart and take all the fun out of it.

     

    reply to this | link to this | view in chronology ]

    •  
      identicon
      dorpus, Jul 24th, 2007 @ 7:04am

      Re:

      Supercomputers are already being used for cancer research. Research into games is useful for validating the claims of advanced players -- their theories may or may not be true. A computer may also discover patterns that no human player had ever conceived. We've come a long way from the 1980s when chess grandmasters used to claim that computers were "incapable" of ever beating a human grandmaster, because they lacked a mystical human "intuition". (The same grandmasters also used to claim that women are incapable of becoming grandmasters because they lack an Oedipal drive.)

       

      reply to this | link to this | view in chronology ]

    •  
      identicon
      Strofcon, Jul 24th, 2007 @ 10:12am

      Re:

      It seems to me that the important thing to take from this article is the idea that they are working on computer "intuition." (Most people would associate that with artificial intelligence, by the way, in case you hadn't kept up.)

      The benefits of making machines that can out-think humans and (supposedly) also mimic our ability to "feel" and reason (the things we seem to claim set us apart from machines) are huge. If a machine can out-compute and out-think us, as well as have some human-like learning/intuitive capacity, it could very easily lead to answers to medical questions that would take us many years to find on our own.

       

      reply to this | link to this | view in chronology ]


Add Your Comment

Have a Techdirt Account? Sign in now. Want one? Register here
Get Techdirt’s Daily Email
Save me a cookie
  • Note: A CRLF will be replaced by a break tag (<br>), all other allowable HTML will remain intact
  • Allowed HTML Tags: <b> <i> <a> <em> <br> <strong> <blockquote> <hr> <tt>
Follow Techdirt
A word from our sponsors...
Essential Reading
Techdirt Reading List
Techdirt Insider Chat
A word from our sponsors...
Recent Stories
A word from our sponsors...

Close

Email This