There are more possible moves in a game of chess than there are atoms in the known universe. So how do computers, which are officially better chess players than humans now, know which moves to make and which to avoid? The Thinking Machine visualizes the thought process of a simple chess computer, as it traces its way through all of the possible moves it can make in a given game in real time.
Very roughly speaking, there are two objective ways of seeing who is winning a game of chess at any given time. The first is the numeric value of the pieces left on the board, where a pawn is worth 1 point, a knight 3 points, a queen 9 points, and so on. More important than points, though, is board position. Every chess piece on a board controls a certain range of squares, in that an opponent can’t move to those squares without the threat of being captured.
So when computers think about chess, they tend to measure their performance against these metrics. They want to increase the relative value of their pieces compared to their opponent, while dominating the largest area of the board. Where it gets complicated is that they don’t necessarily want to do it just for the next move, but three moves, five moves, and even 10 moves down the line. Computers literally need to think in multiple dimensions to track all the possible outcomes of even the simplest move.
You can see this at work in the Thinking Machine. The board is constantly pulsing with dark and light checkerboards showing which squares the computer sees as being most strongly controlled by which player. For example, if the square d4 is pulsing a dark checkerboard pattern, the Thinking Machine thinks that black has the biggest influence on that square in all the multiverses of moves it has plotted out so far.
When you make a move in the Thinking Machine, you can also see how the computer is thinking, not just about the move it will make next, but the move you will make in response. Orange lines traced between squares and pieces represent what the computer thinks black’s move should be, green lines represent what the computer thinks white’s responding move would be. As these lines overlay each other, they become brighter, representing what the Thinking Machine believes white and black’s best moves are. Ultimately, the Thinking Machine tends to make moves along the paths of bright orange lines (black’s best move, averaged across thousands of possible chess games), as long as they aren’t connected to bright green lines (white’s best move, in the same multiverse of possible moves.)
This all sounds extremely complicated, but believe it or not, this is actually an incredibly simple model of how chess computers work, based upon techniques pioneered over 60 years ago. Today’s chess supercomputers are even more powerful. Try to visualize what’s going on in their silicon cortex, and your screen might explode.