mrd (02/09/83)
The whole debate in net.ai about consciousness seems rather ill-founded. While I certainly don't expect anyone talking about this subject to "define his terms", people have meant many, many different things by the word "consciousness", and so there is a lot of room for argument between people using different definitions. My interpretation of P. Devanbu's use of the word "consciousness" is that he refers to subjective experience of events (physical or mental) as separated from any actual effect of the events. I would have to agree that there is very little one can say about consciousness in this sense - it is a "non-linguistic" experience. This does not mean that we can ignore it or that the fact of our postulating that others have consciousness does not matter. It simply means that we cannot directly analyze it. On the other hand, there are many things which seem closely related to our subjective experience that can be described and analyzed at least to some extent. We speak of states of consciousness, consciousness raising, the conscious and unconscious minds, and so forth. Clearly these are important determinants of one's subjective experience, and just as clearly they are not subjective experience. There is no obvious reason why we cannot study (or even simulate!) these aspects of experience. AI is of course nowhere near this point, and it will take a while just to understand the cognitive aspects of the mind. As to whether subjective experience is derived from these describable aspects of experience - I haven't heard anything very convincing one way or the other. It does look like the reductionists (not necessarily all the way to the physical level) will be making most of the scientific progress for some time to come. This doesn't necessarily make them right. I would like to shift this debate to more analyzable aspects of consciousness. Let's hear some interesting ideas along these lines! My own current favorite is Julian Jaynes' theory of consciousness. His "definition" is complicated, and only peripherally related to AI, so I will not discuss it here. Read his book though if you get a chance; it contains some of the most interesting ideas on the subject around. Mike Douglas decvax!genradbo!wjh12!mrd