Thursday, August 10, 2006

Fuzzy Logic and Bayesian Networks

For many many years now, I've been obsessed with understanding how our memory works, basically how we store things, how we retrieve them and the whole triggering process that supports decision making, etc.
In 1994, I was traveling in a car with some friends across the US. Somewhere in the way from Texas to South Dakota I catch-ed myself thinking on the way we make decisions. In university I studied binary logic (true, false) and three values logic (true, false, unknown). I couldn't believe it, all I studied couldn't be farther away from reality. When we make decisions we use probabilistic calculations, we have a range of options from the ones that are known as facts to others that are just a very slim possibility and we operate with multiple clauses with different percentages of assertion. Luckily I was in the US when I had this thought and less than an hour apart I was in one of this huge american bookstores (no internet at that time ;) where I found this book about fuzzy logic that explained exactly how the logic of n-values worked. Fuzzy logic didn't go far on the practical side of Artificial Intelligence, although when I think about it Bayesian Networks are kind of a "modern" implementation of Fuzzy Logic.
Whenever we implement computer systems that involve some learning processes as well as some decision making capabilities we need to look inside our mind and figure out how it works. Then, the computer resolution of the same issue doesn't necessarily have to be exactly the same as ours, but it'll be a start.
Marvin Minsky with the society of the mind is a great attempt to achieve this analysis. He defines all this little agents that operate in our mind which is very computer oriented as an architecture.

No comments: