What I learn from chess and computers

I take these points to be a jumping off place for thinking about computers and future economic growth, and wages, more generally.  The AI revolution basically came first to chess!  Of course chess is sustained by a mix of donations, corporate and political sponsorship, wage labor (e.g., lessons), and volunteer labor, so it is hardly a metaphor for the economy as a whole; still we can see how computer labor and human labor might fit together:

1. Databases equalize preparation opportunities for the top players.  Those who rise to the very top have very strong creative skills.  In relative terms, being a chess “grind” is worth less than in times past.

2. If the computer is set at 2200 strength, “me plus the computer” (I override it every now and then) almost always beats “the computer alone.”  Often we beat “the computer alone” very badly.  If the computer is set at full strength, my counsel is worth much less, although it is not valueless.

3. With a computer set at full strength, the useful “team” requires a much stronger human team member than I.  The required education level — for the team’s “wage premium” — is ratcheted up.

4. Chess is an area where educational reform has been extremely rapid and extremely successful.  Chess education today revolves around learning how to learn from the computer, and this change has come within the last ten to fifteen years.  No intermediaries were able to prevent it or slow it down.  Humans now teach themselves how to team with computers, and the leading human players have to be very good at this.  The computers which most successfully team with humans are those which replicate most rapidly.

5. There are many more chess prodigies than ever before, and they mature at a more rapid pace.

6. We used to think that computers would play chess like we did, only “without the mistakes.”  We now know that playing without the mistakes involves a very different style from what we had imagined.  A lot of human positional intuitions are garbage, and the computer can make sense out of ugly-looking moves.  A lot of the human progress since then has involved unlearning previous positional rules and realizing how contingent they are.  Younger players, who grew up playing chess with computers, are especially good at this.  For older players, it is a good way to learn how unreliable your intuitions can be.

7. Highly exact and concrete analysis, and calculation of variations, is now the centerpiece of grandmaster chess at top levels.  We have learned how to become more like the computers.  The computers have taught us well.

8. Chess-playing computers still are not meta-rational.  They do not understand what they do not understand very well, for instance blocked positions and long sequences of repetition.  That is one reason why human-computer teams are so important and so productive.

Here is Kasparov on Watson.  Here is Kasparov on AI and chess.  Here is a good treatment of human-computer teams.


Comments for this post are closed