harry@aeshq.UUCP (Harry Pulley) (09/21/90)
Computo, ergo sum? [I cannot recall who originally quoted this...and I don't mean Descartes...] I think that there is a missing component in the definition of an intelligent system. This may even be a concept which will separate humans from cats, etc. I believe that an intelligent system must be able to 'think'. I do not feel that thinking is a process which is entirely covered by reasoning. To think is not to know. A computer knows what the answer to a mathmatical function is. It knows what the outcome of a predictable system will be. But to think, a system must be able to examine a situation and say "I think that this outcome will occur, but I am not sure." Humans can observe an event, not knowing all factors which are involved, and make a decision regarding the outcome of the event. The important point here, is that the decision may be wrong. An intelligent system must be able to take all things it knows, and be able for formulate something which is new. It must be able to examine a problem, try all known solutions, and then think of a new solution if it exhausts all proven methods, unsuccessfully. I think that this, more than anything, will be a major obstacle on the road to creating a truly intelligent system. HCPIV
djones@megatest.UUCP (Dave Jones) (09/29/90)
From article <772@aeshq.UUCP>, by harry@aeshq.UUCP (Harry Pulley): > [ speaking of "intellegent systems" ] > ... But to think, a system must be able to examine a situation and say > "I think that this outcome will occur, but I am not sure." Of course, probability theory deals with exactly this kind of thing. There is also a new branch of computer science, closely related to probability, called "fuzzy logic", a.k.a "fuzzy-set theory". It is still a new discipline, but it looks very interesting and promising. In standard Boolean algebra, the "characteristic function" of a predicate maps objects into the set {0,1}. If the predicate is satisfied for that object, it maps to one, otherwise it maps to zero. Fuzzy predicates map objects into the interval [0,1] not to 0 or 1. The number indicates to what extent the predicate is applicable to the object. The mapping is usually somewhat arbitrary. For example, a 5'11'' person might have a "tall" number of .8, while a seven-footer would be considered completely "tall" at 1.0. Many of the operations and theorems of Boolean algebra can be abstracted in the theory of fuzzy logic.