• Question: What is your favourite theory of AI and what they might evolve into

    • Keywords:
    Asked by camd0g2007 to William, Nick, Maja, Giuseppe, Diana, Andrew on 30 Jan 2020.
    • Photo: Giuseppe Cotugno

      Giuseppe Cotugno answered on 30 Jan 2020:

      A topic I always liked in AI is fuzzy logic.

      Logic is the discipline which is used to manipulate truth values and make conclusions. For example: Socrates is a man, all men are mortal therefore Socrates is mortal. This is a logical conclusion. Logic is the basement of the the first generation of AI systems, the expert systems. Those systems were modelling the knowledge of a human (the expert) and were proven to be valid in diagnosing diseases and other things. The problem is that those systems are too stiff because for them it is all a black and white e.g. either the water is poisonous or not, he can’t be a little bit drinkable and a little poisonous for example.

      Fuzzy logic overcomes this limitation. It provides a way to formally define mathematically the gray areas. You can have water that is rather dirty and a little salty because there is a mathematical definition (a function) for “rather” and “a little”. Fuzzy logic those days is really the underdog of AI. It was discovered long ago, it was used a little bit and then it was put aside. I think that this formalism has a lot of potential and it might resurrect first generation AI by producing less stiff and more human friendly systems.