“Brains Vs. Artificial Intelligence,”
In what has been dubbed “Brains Vs. Artificial Intelligence,” for the next two weeks four of the world’s best Heads-Up, No-Limit, Texas Hold ’em poker players are taking on a highly advanced computer program designed by researchers at Carnegie Mellon University for their share of a $100,000 prize pot.
From 11 a.m. to 10 p.m. every day till May 7, with one day off for rest, they will play 80,000 total hands of poker in an effort by CMU’s researchers to prove for the first time that a computer can beat humans at this most complex version of poker.
.
.
.
The public is invited to watch the event in person at Rivers Casino for free. Visit here to check out the daily progress and watch each competitor hand-by-hand.
The full story on link.
http://www.post-gazette.com/local/city/2015/04/24/CMU-Brains-Vs-Artificial-Intelligence-poker-tournament-begins-Pittsburgh-s-Rivers-Casino/stories/201504240281
Quote: zoobrewThe humans took the early lead against the computer this morning at Rivers Casino.
The "humans" programmed the computer.
Humans are the most intelligent animal on the planet. But, that's not saying much.
AI can be encouraged to make the proper move in any scenario with no emotional entanglements, no second guesses, no reserves, etc.... In the long run a perfect AI program will beat any human at anything. This is simply reality. Humans are too ingrained to emotions and dispositions which can be and likely will be our downfall.
For many years already there have been AI programs that can beat the best Masters at chess. Poker will only be a matter of time. And, life in general will be only a bit longer still.
Quote: GandlerAI will win. Even if not in this event, then eventually.
Humans are the most intelligent animal on the planet. But, that's not saying much.
AI can be encouraged to make the proper move in any scenario with no emotional entanglements, no second guesses, no reserves, etc.... In the long run a perfect AI program will beat any human at anything. This is simply reality. Humans are too ingrained to emotions and dispositions which can be and likely will be our downfall.
For many years already there have been AI programs that can beat the best Masters at chess. Poker will only be a matter of time. And, life in general will be only a bit longer still.
Heh, this is basically the premise of the Matrix.
NLHE is not the "most complex version of poker." It is among the least. And the article even stumbles (apparently unknowingly) into a mention of the fundamental flaw in the basic structure that is a primary reason:Quote:at this most complex version of poker.
Quote:In particular, Claudico fairly regularly bet “all-in” when it felt like it had a really good hand,...
Quote:Two-player no-limit Hold’em, Sandholm said, has 10161 (1 followed by 161 zeroes) situations, or information sets, that a player may face —vastly more than all of the atoms in the universe. By contrast, the easier game of limit Hold’em, in which bets and raises are limited to a pre-determined amount, has only 1013 (1 followed by 13 zeroes) information sets.
http://www.cmu.edu/news/stories/archives/2015/april/computer-faces-poker-pros.html
I started a thread earlier.
Here's a question for the readers: how many computer programming languages besides BASIC use ^ for exponentiation? I remember FORTRAN uses **, and C requires a separate function (pow( ), IIRC). It just seems strange that this is what caught on in text.
Quote: ThatDonGuyif you are going to express something raised to a power, don't just smash the numbers together.
I'm sure it was an oversight of the copy & paste process. Sometimes, some of those quirky formattings get mangled.
Quote: ThatDonGuyHere's a question for the readers: how many computer programming languages besides BASIC use ^ for exponentiation? I remember FORTRAN uses **, and C requires a separate function (pow( ), IIRC). It just seems strange that this is what caught on in text.
http://rosettacode.org/wiki/Exponentiation_operator
Oh my. That is an extremely simplistic misunderstanding that's really not at all relevant to the comparative complexity of the game, because most of those will NOT be "faced" among competent non-random players. Here is what most commonly will be: one or two streets, all-in, call or fold, game over. For much more depth on that topic, I'd refer you to the extensive writing of Ed Miller on that subject in both short article form from poker publications available online, and treated in more depth by him in some of his books that are widely regarded as among the fundamental texts for the game.Quote:has 10161 (1 followed by 161 zeroes) situations, or information sets, that a player may face
Perhaps it requires more lines of computer code when programming for each theoretical possible "information set" as defined above. But most of them will not occur, and that silly metric is not = comparative complexity of decision making. I'd much rather play an AI device in a NL structure than the LHE machines that currently exist in some casinos. But I wouldn't expect to win from either without being a superb HU player, which I am certainly not.
Quote: DrawingDeadOh my. That is an extremely simplistic misunderstanding that's really not at all relevant to the comparative complexity of the game, because most of those will NOT be "faced" among competent non-random players. Here is what most commonly will be: one or two streets, all-in, call or fold, game over. For much more depth on that topic, I'd refer you to the extensive writing of Ed Miller on that subject in both short article form from poker publications available online, and treated in more depth by him in some of his books that are widely regarded as among the fundamental texts for the game.
I'd much rather play an AI device in a NL structure than the LHE machines that currently exist in some casinos. But I wouldn't expect to win from either without being a superb HU player, which I am certainly not.
I think it's similar to all the chess moves that would be obviously not preferred as a next move, but still a computer needs some sort of sorting mechanism in it algorithms in order to not just be brute force calculation.
It's why a checker game is solved and a chess game is not. All those non-practical moves need to be eliminated in one manner or another, so I guess you could refer to it as complexity.
I think I get that, and I gather the "complexity" remark was from the particular point of view of the distinct task of the programming geeks. And for programming purposes rather than playing, perhaps that's right. Or not, I don't know.Quote: rxwineI think it's similar to all the chess moves that would be obviously not preferred as a next move, but still a computer needs some sort of sorting mechanism in it algorithms in order to not just be brute force calculation.
It's why a checker game is solved and a chess game is not. All those non-practical moves need to be eliminated in one manner or another, so I guess you could refer to it as complexity.
But the result of particularly using a NLHE structure will not in the end produce a relative chess match as played compared to checkers in choosing another commonly used poker game format, unless it is a result of the opponent being Mr. Checkers or Ms. Tic-Tac-Toe, and the AI executing a player-specific learning function. Try playing the algorithm in the LHE machines locally at Aria or Red Rock or Caesars Palace Bellagio over a significant number of hands. And record the cards and action in the hands (you can view HAL9000's cards when the hand is over). I have, and ended up with a hugemongous spreadsheet from doing so after finally fleeing HAL's pod bay doors weeks later. It is not simply a straightforward matter of "if this, then that" over anything like a smallish set of realistic potential datasets. For one thing you'll find that HAL's play in identical circumstances will vary, as it should.
In play, it is not checkers, nor is it chess, in either case.
EDIT: Corrected a location of the HAL9000 Pokah Wizard thingamajigger.
Chess is nothing like poker. There's to many variables in NL. Bluffing, bet sizes, slow playing, tilting, trapping, players mood,style of play, emotions.Quote: GandlerAI will win. Even if not in this event, then eventually.
Humans are the most intelligent animal on the planet. But, that's not saying much.
AI can be encouraged to make the proper move in any scenario with no emotional entanglements, no second guesses, no reserves, etc.... In the long run a perfect AI program will beat any human at anything. This is simply reality. Humans are too ingrained to emotions and dispositions which can be and likely will be our downfall.
For many years already there have been AI programs that can beat the best Masters at chess. Poker will only be a matter of time. And, life in general will be only a bit longer still.
You have to teach the computer to read minds or predict what a player will do or detect unique betting patterns or emotions. That will be very hard to do. for the exact reason you think it will be good. It cant detect emotions nor does it have any. Sometimes Humans change their mood (even if slightly) each second for hundreds of different reasons.
I think it will be a very very long time before they could put something like this (NL) in a casino environment.
I watched the actual play for quite a while and it's very weak especially free flop. It's bluffing was horrible, but it got bluffed often.
What I would like to see done. Without the computer or human knowing. Both the human and computer get the exact same starting hands same positions, flop, turn and river. Take all luck out of the situation. This should help define the computers skill quickly.
Quote: AxelWolfChess is nothing like poker. There's to many variables in NL. Bluffing, bet sizes, slow playing, tilting, trapping, players mood,style of play, emotions.
You have to teach the computer to read minds or predict what a player will do or detect unique betting patterns or emotions. That will be very hard to do. for the exact reason you think it will be good. It cant detect emotions nor does it have any. Sometimes Humans change their mood (even if slightly) each second for hundreds of different reasons.
I think it will be a very very long time before they could put something like this (NL) in a casino environment.
I watched the actual play for quite a while and it's very weak especially free flop. It's bluffing was horrible, but it got bluffed often.
What I would like to see done. Without the computer or human knowing. Both the human and computer get the exact same starting hands same positions, flop, turn and river. Take all luck out of the situation. This should help define the computers skill quickly.
Programs can detect emotions, facial expressions, it won't be long before a program can keep track of facial ques and betting styles to predict bluffing likleyhood. Or even just betting patterns.
It has no emotions, but it can be trained to bluff, especially in ideal scenarios (based on past plays of players). It can be set to evaluate the betting styles of players and their frustration, and bluff accordingly based on past plays.
I agree its far different than chess. But I was just making the point, at any specific task a computer can usually be made to be better because it can instantly see all of the data and automatically know the ideal play every turn.
You experiment would be interesting. But I feel that after a couple hand both the computer and the players would realize what was happening (at least I would hope so). You would maybe have to integrate it in a single hand basis here and there throughout the game to see how their decisions compare to fair/random hands.
I agree with this part.Quote: GandlerPrograms can detect emotions, facial expressions, it won't be long before a program can keep track of facial ques and betting styles to predict bluffing likleyhood. Or even just betting patterns.
Especially if it started reading heart rhythms, body temperature, breathing patterns, sweating, micro eye movements, body quirks, etc.
Things most humans can't do. I'm assuming the poker community would never allow that. It takes away from what poker really is. I Assume at some point computers will be able to actually read peoples minds, but that's ridiculous to think we would play poker with something that could do that.
Assuming a computer won't have the ability to read your mind or detect the things I mentioned from a different location. I don't think it would win in an online situation playing a good player. Especially if it randomly played vs good players.
I don't think they will make a NL cash game computer AI that could profit playing good players.
Quote:Though three of the four pros had higher winnings than Claudico, their $732,713 collective lead over the A.I. program was not quite large enough to attain statistical significance — in other words, the results can't be accepted as scientifically reliable. In all, $170 million was "bet" during the two-week "Brains Vs. Artificial Intelligence" exhibition. So despite the apparent lead by the humans, the competition ended in a statistical tie.
Quote:In the final chip tally, Bjorn Li had an individual chip total of $529,033, Doug Polk had $213,671 and Dong Kim had $70, 491. Jason Les trailed Claudico by $80,482. Each of the players is ranked among the world's top 10 professionals in heads-up (two-player) no-limit Texas Hold'em.
http://www.scs.cmu.edu/news/poker-pros-rake-more-chips-carnegie-mellons-artificial-intelligence-program-during-80000-hand-contest
Quote: GandlerAI can be encouraged to make the proper move in any scenario with no emotional entanglements, no second guesses, no reserves, etc.... In the long run a perfect AI program will beat any human at anything. This is simply reality. Humans are too ingrained to emotions and dispositions which can be and likely will be our downfall.
You forgot the most important thing: Even if Humans would know perfect (mixed) strategies, they would still lose.
AI can draw practically unbiased random numbers. Humans fail at this quite miserable.
