Friday, September 14, 2018

Suppose you have to be smarter to drive than be a chess champ?


Chess is an informal IQ test. A goal of AI was to design a computer that could outperform the best chess-players. That was a way of proving AI. Mind you, calling what computers do "intelligent" is Pickwickian. 

But be that as it may, what if it takes more intelligence to perform a mundane task like driving than playing chess? If we can design computers to beat even the best players at chess, but we can't design automated cars to reliably do what human drivers can do, then that presents a paradox for AI. 

This may go to the fact that in chess, the variables are quantifiable and predictable, while driving conditions are not. In that context, human intelligence is adaptable in a way that AI is not. 

1 comment:

  1. Yeah, with chess, there are basically six different kinds of pieces confined to an 8x8 space. It’s a lot of calculation for a human, but once computers learned a few other (sound) rules, they can pretty much understand the game comprehensively.

    ReplyDelete