Computers are good at chess, mostly

If you want a chance at beating a computer at chess, there is one simple way: Make the computer itself move the pieces across the chess board. For all the playing strength of the programs, the computer-driven robots that move the pieces don’t do such a great job.

From Tyler Cowen’s fascinating Average Is Over, about chess, technology and the hollowing of the American middle class.


  1. Down the rabbit hole! Cool Youtube videos of chess robots, for sure, but orthogonal to the point that Dr. Cowen’s making, or perhaps reinforcing it.

    The quote is excerpted from a longer discussion of the regularized environments needed for the current generation of robotics to function well. Your hobbyist (and physical chess robots in general) seem to have solved the problem of moving a piece in a tightly controlled environment – the specific board and specific chess pieces for which it’s been designed. Good for them! Dr. Cowen is arguing that the problem has not yet been solved for the general case. Having a robot succeed with an arbitrary chess board and pieces of a type that it hasn’t seen yet is still a tough problem. Not so much for the human. “Complicated environments – also known as the real world (even real chessboards) – are tricky.” One of the things we do as a result (and this is the deeper point he’s making) is that we regularize environments where we can and let machines handle the task, and turn to humans for those activities that can’t be regularized. That’s why, for example, in human-computer chess, we dispense with the physical board entirely.

    Watching robot videos on Youtube (not just chess, as I said, down the rabbit hole, and thanks for that!) very much reminds me of a joke I once read about a dancing bear.

Comments are closed.