I just watched a show on CBS about A.I. and Texas Holdem poker.
Games are thought to be one of the most difficult things for A.I. to master because of the various strategies involved, particularly that of bluffing. Some history to acquaint you with:
Chess was the first pinnacle to be mastered about twenty-odd years ago when the best grandmaster in the world was finally easily defeated by Big Blue — IBM`s mainstay computer.
About 10 years ago, Go, which was then considered the most difficult board game in the world and one that it was thought a computer would take many, many more years to solve, beat the world’s best grandmaster.
And now Texas Holdem has been conquered! Once again the timeline has been compressed for the speed of A.I. learning. Texas Holdem, for those of you not familiar, is standard poker that in order to win over time includes a large element of not only luck of the draw of the cards, but what has always been thought to be the key ingredient of bluffing that sets it far above chess and Go.
Why? Because with chess and Go A.I. simply had to learn to win by playing against itself million and millions of time to figure out all possible moves. Humans are not able to make those almost infinite number of moves to learn the perfect game.
For the game of Texas Holdem, A.I. not only had to learn to play against itself an almost infinite number of times to learn the probabilities of the possible combination of poker hands, but it now had to learn what the presumed human element of bluffing was.
A.I. then taught itself all there was to know about bluffing. It has mastered one of the last thought to be unique elements of humankind.
My summation is that in short order A.I. has exhibited exponential growth. It is truly scary to me. There is nothing that can stop this progress as A.I. continues to teach itself to out-think us all.
