Functionalism and AI

Posted on August 6, 2010
Filed Under Philosophy | Leave a Comment

Chris and Gary’s disagreement in the comments of my brief post on functionalism got me thinking. Now I’m pretty skeptical about functionalism because I just can’t see anyway to make 3rd person into 1st person logically. (Peirce sees his three categories as irreducible and provides some interesting arguments for that which I won’t go into) All that said many people embracing functionalism aren’t bothered about the issue of qualia because in a certain sense it is irrelevant for their position. I came upon a pretty compelling argument for this although I honestly don’t know the origin of the argument. One of the strongest arguments against functionalism goes back I believe to Searle originally. He notes the difference between a simulation of a motor and a real motor. The real motor depends upon the substances involved in a way that a simulation doesn’t. The real motor burns the gas to produce torque but a simulation can’t and only produces a representation of torque rather than torque itself. So relative to consciousness, the argument goes, functionalism can at best produce the representation of consciousness and not consciousness itself.

The AI proponent has a pretty good response though. They note that what they are after only is representations. What they are concerned with is speech and writing as the output of a conscious mind and not consciousness as such. The output of a person that is relevant are our words and writing while the output of a simulation will be the same. Thus the outputs are the same.

This is pretty compelling to me. It avoids the central question of how to produce this output of course. It may well be that one simply can’t generate the function without the underlying substances. Think, for instance, chaos theory where in some circumstances even minuscule differences between reality and the model lead quickly to huge functional differences. The debate then becomes over how to model the brain and where instabilities matter. At this stage though our knowledge of how the brain functions is rudimentary enough so as to still allow a pretty strong skepticism of AI.

Effectively then the main skepticism of functionalism is (a) it leaves out key aspects of what we call consciousness or mind and (b) one can’t model the substances as a practical matter. I think most of the skepticism is really about the limits of digital simulation, but that’s a bit unfair to apply to functionalism as an ontological theory. I think the ontological and practical issues get conflated a lot.

For the record I’m pretty skeptical about functionalism as an ontological thesis but primarily because I think it leaves out too much. I’m really skeptical about AI just because I think as a practical matter intelligence and consciousness are just far more complex than AI proponents tend to believe. (Which isn’t to say AI research can’t be useful for other reasons)

Related posts:

  1. Gary and Peirce on Mind and Functionalism
  2. Dust Argument
  3. Peirce and Consciousness
  4. Consciousness
  5. Vallicella on Consciousness
  6. Consciousness Without Attention?

Comments

Leave a Reply