Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!tdatirv!sarima From: sarima@tdatirv.UUCP (Stanley Friesen) Newsgroups: comp.ai.philosophy Subject: Re: Emergent Properties Keywords: chaos, science, prediction Message-ID: <30@tdatirv.UUCP> Date: 14 Oct 90 15:58:53 GMT References: <1990Oct12.214636.7945@ncsuvx.ncsu.edu> Reply-To: sarima@tdatirv.UUCP (Stanley Friesen) Organization: Teradata Corp., Irvine Lines: 130 In article <1990Oct12.214636.7945@ncsuvx.ncsu.edu> fostel@eos.ncsu.edu (Gary Fostel) writes: > I suspect that what is and is not going to fall under many peoples > notion of "emergent" is going to depend on the level of understanding > of the people and the moment. The more easily a property can be > predicted using availble methods, the less likely it is to be an > "emergent property". ... > but I wonder > if as many would be comfortable with the statement that a tabular > printout is an emergent property of a particular set of Cobol > statements. After all, the table is not at all readily predicted > from any one of the cobol statements. I would have no problem with this. The operation of any computer program is an emergent property of the individual instructions that compose it. This is because i t is the *organization* of the instructions that determine the program's behavior, not the set of instructions themselves. (There are many different programs that can be constructed from any given collection of instructions, so the program behavior is not even theoretically predictable from the indiidual instructions). And since the organization of the instructions is a global property of the entire program, there is no lower level at which the total behavior can be characterized, thus the behavior is emergent. > A more interesting (to me) issue is whether there might be some > properties that really are VERY hard to predict or model based on > the constituents. For example, non-linear dynamical systems are > often essentially impossible to predict, not due to lack of theory > but for intrinsic reasons -- so called chaos theory. This is not related to emergence at all. It is a matter of computational hardness. The full properties of a chaotic system are inherent in the simplest description of the system, they are simply unrecoverable. Thus no *new* properties are produced at 'higher levels', the existing properties are simply made partially visible. This *is* an important question. It is the basic reason why pure deductive reasoning is of limited value in the real world. It is why all living things use heuristic analysis of some sort to produce the merely *probable* rather than the *certain* result of traditional computation theory. This suggests that any 'real' artificial intelligence will be prone to error just like we are. It also means that there will always be a place for 'normal', deterministic, non-AI computation. > In the domain > of artificial intelligence (or perhaps right outside it :-) are > some people who argue that human intelligence can not be duplicated > or modeled because of the subtle but undeniable infusion of EVERY > detail of life into the decisions and thoughts of a moment. If > memory serves me, Penfield is a recent example of this group. Even if he is right, I do not see why we cannot design a software system to do the same thing. It would be very difficult, and it would take computers that make Crays look like home computers to do it in real time, but it should be possible. So a complete cross-indexed associative memory would be needed, so all decisions would have to be cross-checked with the memory, so all events of concern to the computer would have to be digested, understood and indexed. This is *not* impossible. My main problem with most of these people is that they take differences between mental processes and *current* computer technology and treat these as intrinsic limnitation of computation. *BULL*! So far no mental process that is well-understood cannot be simulated in a computer with sufficient power. I see no reason why the ones we do not understand should be any different. > An interesting property, "P", that systems might have, would be > that they produce behaviors that are drawn from a well defined > set of behaviors, even though the direct prediction of which > behavior is intrinsically intractable. > a set of neurons (esp real neurons) might not > be so easy to predict. Neural "programming" is more a question > selecting alternative neural nets from a set of possible nets until > the bahavior happens to be the one desired. Such a system might > well have property P if it could be shown that the selection > strategy was really the only way to get the desired behavior > with probability 1. You seem to be assuming that there is exactly one desired behavior, as if intelligence required exactly matched behavior! So a computer intelligence would make a different decision than I would, it does not matter if the decision was arrived at using 'intelligent' processes. That is intelligence is a broad *range* of systems, and the exact duplication of any one of them is unnecessary to generate intelligence. So all that would be necessary is that the computer intelligence be based on a chotic system that is *similar* to the one humans use. And deterministic computer programs can produce true chaotic behavior, my lock screen on my workstation is an example. > The selection stratgy for a neural net may bother some folks, > who feel they "design" nets. In the case of artificial nets, it > is probably true that the net can be apriori "designed" and then > built. I would say that those systems do not have property P. > Natural nets and some synthetic nets, are often "trained" > which really means that a sequence of nets are produced, with the > sequence terminating when a net with the desired behavior is > found. You seem to be treating identically wired nets with differing weights as if they were different nets. I think this is probably not a useful approach. Since in real neural systems the weights are constantly changing in response to experience, this would lead to the strange result that my brain today is a different set of nets than it was last year! [Remeber, the basis of memory in living neural systems is the changing of the connection strengths]. > Perhaps my property P is what others are calling "emergence" and > I am just befuddled, or perhaps P is something else and I'm still > befuddled anyway. I would say that P is the property of unpredictability, or intractibility. It is only superficially similar to emergence, which *can* be predictable if the structure of the system as a whole is taken into account. > If you would like to spend some time sharing > my befuddlement, consider whther there is a relationship between > systems with property P and problems which are NP complete. In > each case it seems that there is not a way to "get inside" the > problem, and search may be the only way to go. They do seem similar. And it is this inability to 'get inside' that makes intelligence necessary for decision making. If chaotic and NP complete problems did not abound in nature simple analytic logic would always produce the right answer, and model-based search strategies would not be needed. Was I any help????? Thanks for listening -- --------------- uunet!tdatirv!sarima (Stanley Friesen)