Poker is like real life. You have to make decisions on the basis of imperfect knowledge and you often have to bluff in order to achieve your goals. One might think that computers cannot do these things, or at least do these things as well as humans can. On July 23 and 24, we might find out whether this is the case.
On July 23 and 24, the poker-playing-computer named Polaris, built by the AI computer scientists at University of Alberta, will play Texas Hold’em against Phil (aka the Unabomber) Laak and Ali Eslami for $5000. What is especially interesting about this match is that it has been designed to eliminate luck as a factor.
How do the scientists plan to eliminate luck as a factor? In one room Polaris will play Laak and in another room Polaris will play Eslami. Polaris will get Laak’s cards when playing Eslami and will get Eslami’s cards when playing Laak.
If Polaris wins the poker match, this will have interesting philosophical implications. Particularly, it will remove another barrier from seeing that computers could think.
Computer science has already shown that computers can “go beyond” what their programmers programmed them to do, when in 1997 Deep Blue, the IBM chess-playing computer defeated world champion Gary Kasparov. The fact that Deep Blue beat Gary Kasparov (Kasparov complained that Deep Blue cheated) did not show that a computer can produce new knowledge. Rather, Deep Blue’s analysis of the game showed this, because it showed that certain endgames that always had been considered draws by humans were in fact wins.
If scientists can make computers that can come up with new knowledge, that can make good decisions on imperfect knowledge, that can misrepresent their knowledge, then they are one step closer to making a thinking mind.