Tuesday, 6 March 2012

Bye bye AI

I'm a bit sad because today I hand in my Artificial Intelligence coursework, and this is probably the last time I get to use the logic programming language Prolog. It's a very cool language, essentially made of punctuation rather than words. I still have a few more AI lectures left though. We just had an interesting one about the problems with using Classical Logic for actual reasoning. In classical logic, if you have a set of things that are true, and another thing, let's call it P, follows from that set, then, whatsoever you may add to that set, P still has to be true. But we need a different sort of logic for practical reasoning, a defeasible or default logic. In this way of proceeding you have a default, e.g. birds can fly, and then you assume the default that if Tweety is a bird Tweety flies. When you find out that Tweety has a broken wing, or Tweety is a penguin, or Tweety is dead, you change that default. In classical logic you could only do that by stating that birds fly if they are not dead and don't have a broken wing, and aren't a penguin or an ostrich, and aren't glued to the ground, etc, etc, and then when you had Eddie (an eagle) you could only prove that he was able to fly if you could also prove that he wasn't dead, glued to the ground, unwell, an ostrich, etc, etc. Default reasoning is like the legal position where you assume that an accused is innocent until proven guilty. It's also known as the Closed World scenario, like in databases, where it's assumed that what is not recorded is not true. In Prolog it's called Negation By Failure. If you can't prove X then you act like for practical purposes you have proved not X.

This seems very interesting to me. It seems like the default reasoning method is a bit like the whole quantum, Schrodinger's cat thing, where only the observable is true. Does it make much difference to people's personalities and attitudes to spend their entire time in an environment where only the observable is true? This seems like almost the opposite of the Humanities approach, and I think this is what Imperial is lacking by having no Humanities here, just a loose association with the Royal College of Art and the Royal College of Music. Tomorrow I may or may not go to a talk by an anthropologist who works on people who work with mice, which seems like an interesting related thing.

No comments:

Post a Comment