How to think about things in a manner which helps to design robots

Why not take a computer which plays chess well by looking at all of the possibilities and have it "sleep"
on the various board positions and relate board positions to "whether the future has positive

For example, humans play chess, lose games and then associate the board positions with "being weak"
or "being strong" for the other player.  People have difficulty when first trying to learn.  After they have
learned, they sometimes can drive for quite awhile without being conscious of it.  In other words, the
human brain has the ability to learn procedures which contain goals and sub-goals and carry them out
in their sub-conscious.  

If people dream about the chess games which they lost or won and incorporate board positions with
positive or negative "feelings", then why can't computers be programmed to do the same thing?  One
would have custom chess position chips such as Deep Blue did.  Then you add the rest of the digital
computer functions like was done with Deep Blue.  The difference would be to incorporate neural
networks which would learn good and bad positions.