Reading prahou's written art (slash prophecy) got me thinking about
robot thoughts. I really need to get around to Do Androids Dream Of
Electric Sheep.
Monte Carlo integration sucks, and is pretty similar to breadth
first searches in this way.
Monte Carlo integration is like playing darts by putting on a blind
fold, spinning yourself around a few times, and throwing darts at
random, but remembering how many times you heard the thud of a dart
hitting a dartboard.
If you are trying to learn about dart boards near you, this does
kind of work as a last resort but it's clearly not great.
A robot might be able to throw darts thousands of times faster than
you, but this doesn't categorically improve the idea.
In contrast, consider the ACL2 waterfall. ACL2 makes promising
substitutions in your theorem, then if it recognises the situation
it finds itself in, inducts that until it resolves an answer, or
gets tired and stops anyway (like if your induction scheme is a big
search).
The moral is not the Capitalist Hero: Hey, a robot that throws
darts, kill the poor immediately- robots don't have rights!
The moral is that blindly firing darts into your surroundings, with
out thought for the ensuing damage is bad.
Making thoughtful, complex and contextual decisions is a good idea.
