cbostrum (01/23/83)
There is a controversy associated with all of the "special sciences" as to whether there is a legitimate "level" of existence that is the object of their study, or whether ultimately they are merely doing some sort of physics, and that their science will only become fully understood or legitimate when the ultimate reduction to physics is produced. There is one classical chain which goes: physics, chemistry, biology, pyschology, social science (economics, polisci, etc), where every element in the chain is supposed to be reducible to the previous one. Without disputing this, I would like to know where people think AI fits in, not to the chain, but just in general. A lot of talk in net.misc previously went on all about making nueral models, and watching them evolve, and this sort of thing. This would imply that there is no signifigant "knowledge level" as newell calls it, and that there is no meat to taking what dennett calls the "intentional stance". Or perhaps it merely implies that those positions are too difficult to get results with. Personally I at least hope (and presently believe) that both implications are false. Surely there are signifigant things about intelligence we can learn without going to the low-level brain stuff, and surely there must be signifigant things we can learn on the "higher" level that would actually be IMPOSSIBLE to learn on the "lower" level. It seems to me that AI is predicated upon these optimistic beliefs. What are others thoughts about this?
tombl (01/25/83)
One of the problems which seems central to the issues CB has raised concerns what is knowable. One may say that chemistry is reducible to the laws of physics according to our understanding of the physical world; it is not, however, possible for us to construct or predict the principles of chemistry solely from a knowledge of some set of physical laws. Perhaps "has not" would be better then "is not", but the problem is in what I perceive to be a limitation on human mental abilities. We could just as well ask the question of whether it is possible to know the OS/360 operating system. Most peoples' answers would be no. Steering clear of the existential dilemma, what we find is a body of scientific theory describing the world at a physical level; also a body describing the world at the chemical level. Then there is the theory which attempts to relate the two. We can find the same pattern reflected in diverse habitats: in the structure of institutions, in the algorithms we employ to solve problems, and in the architecture of computing machines. This is not an argument for or against a coherent scientific explanation of the world based upon laws of physics, but a suggestion of what the structure of our understanding of the it is and will be. Given an estimated 10^10 cells in the brain, with an average fanout of say 100 on input and/or output, and several modes of interaction between "wires" and cells, I must say that the prospects of reasoning about all brain function (especially behavior) from our limited knowledge of its substructure, seem to me rather unpromising in my lifetime. Higher levels of organization, and the notion of several such levels, seem plausible. AI research, at best, attempts to define one or more of these levels. Beyond that, I wouldn't make any great claims (there are entirely enough people doing that), but I certainly am not going to trade the opportunity to find out for the grand opiate, the illusion of a secure job, or for the intrinsic pleasure of attending church every Sunday morning. Tom Blenko decvax!teklabs!tekmdp!tombl ucbvax!teklabs!tekmdp!tombl