AI has radically changed over the last 20 years. So if you look at when Kasparov was playing against Deep Blue in the 1990s. Which was the best AI we could do for chess at the time, basically what the computer was doing was look ahead. So if I go ahead a few moves, what does the new state of the board look like, and then try to evaluate if this is a better position. The new AI is completely different. What it is, is they just have millions and millions of boards that they’ve seen from grandmasters for example, and it’s basically pattern recognition. So I’ve seen this state of the board say 50 times it’s exact board position, and if white makes this move, then they win most of the time. Thus, I’ll make the move that causes white to win most of the time. So it’s probabilistic after you’ve done the pattern matching on the board. So in many ways, the old way was more like how humans thought, But now it’s just pattern recognition, but this is a much stronger. What it requires is millions of examples of data to learn from but if it can get that and get all the information then it is far stronger.
