Getting the Right Answer, Winning the Game

Fascinating philosophical story in Janelle Shane’s book on AI “You Look Like a Thing and I love You”.

Programmers were trying to figure out the optimal strategy in a game of tic-tac-toe with an infinite board, and had programs play each other gazillions of times. It turned out the optimal strategy was to play your second x as far away from the first x as possible. This would cause the machine playing you (you are an AI too in this example) to crash trying to model that huge playing board, and you’d win by forfeit.

It made me think that if you took robots and gave them bodies and asked them to generate the optimal strategy for chess, it might turn out that the optimal strategy is to pick up your pawn, kill your opponent with it, and then place it on the board. Your opponent would forfeit the game due to death.

Is it the optimal strategy really? No! You’d go to prison. And this raises the interesting point that the purpose of chess is not to win. The purpose of chess is to have a good time, build a social relationship with the player, something like that.

And yet the notion that the purpose of chess is not to win is something it’s easy in our society to miss . I think that’s a contingent historical fact. We live in a world where certain people are rewarded for winning games (the game to make money, the game to make the prettiest picture) and are allowed to ignore the actual purpose of these activities.

This is true of test-taking also. People are rewarded for getting the right answer on the test. But sometimes the purpose of the test could be better served by giving a different answer, or no answer, or refusing to take the test, or helping everybody in your class get the right answer, so you all get a break.

It’s a similar illusion that comes from a similar social peculiarity — rewarding people for getting the right answer on tests, rather than living well and helping others to do the same.

It’s a bad thing. It’s an interesting example though of how what seem to be big philosophical talking points (eg “the facts don’t care about your feelings”) are actually philosophical mistakes, that follow from prejudices which follow in turn from contingent, not-so-great features of our society.


3 thoughts on “Getting the Right Answer, Winning the Game

  1. I mostly agree, but I’d explore your robot example a little further. A robot is simpler than a human. In the example, it has only one goal. Human activities typically have multiple goals, both conscious and unconscious. That’s true of all human societies, not just ours.

    Chess is an example. It exercises our intellectual abilities, but it also satisfies our aggressive impulses without the need for violent conflict. It’s a classic case of sublimation. The 20th-century chess master Edward Lasker wrote a book called “Chess for Fun and Chess for Blood” to describe that spirit: https://www.amazon.com/Chess-Fun-Blood-Edward-Lasker/dp/4871871339/

    Our human needs to struggle against obstacles and to “win” are part of our biological makeup. The problem is to satisfy those needs in ways that are at worst benign and at best socially helpful. Usually, we end up with a mix of good and bad.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s