## November 4, 2007

### Computer with common sense? - continued

Filed under: computer vision/machine vision/AI, reviews, rants — Peter @ 6:21 pm

In the last post I asked “If you can teach computer to recognize objects, you can teach it simpler things. How about teaching computer how to add based entirely on feedback?” I was going to write a post about this but the next day I got an opportunity to approach this differently. At Hacker News I read the post How to teach a Bayesian spam filter to play chess. So I presented this (implicit) challenge.

This would be a better experiment because (1) it is simpler and faster, (2) the feedback is unambiguous, (3) the ability to add is verifiable.

Essentially you supply it with all sums of all pairs of numbers from 0 to 99 and then see if it can compute 100+100.

You can see the whole discussion here (my comments are under ‘pixcavator’). Let me give you some highlights.

First, it was educational for me. Books and sites were generously recommended. This one made me laugh:

My personal recommendation on machine learning is ‘Pattern Recognition and Machine Learning’ by Chris Bishop. But you definately do need a solid mathematical background for that.

I read some of what was recommended about Bayesian method and neural networks – they were mostly irrelevant but interesting nonetheless. The discussion helped me to formulate the answer to my own challenge:

Numerically this is easy, symbolically impossible.

The OP did not take up the challenge but a few people responded positively – yes, the problem can be solved. I started to pose ‘naïve’ questions and it was fun to watch people dig themselves deeper and deeper. This one is my favorite (really deep):

Question: So given a function with f(’1′,’1′)=’2′, the computer will figure out that f(’1′,’2′)=’3′, right?

Another one:

Question: Where would the idea of “is bigger than” or “is the following number of” come from if not from the person who creates the network?