On Fri, 2011-08-12 at 16:30 -0700, pmbauer wrote: > +1 > > On Friday, August 12, 2011 3:16:15 PM UTC-7, Sergey Didenko wrote: > BTW, Is there a case when AI self-modifying program is much > more elegant than AI just-data-modifying program? > > I just can't figure out any example when there is a lot of > sense to go the self-modifying route.
Clearly both are equivalent in a Turing sense. If you are 6 foot tall but "modify that in data" so you are 7 foot tall and always report yourself as 7 foot tall then there is no way to distinguish that from the records. Yet there is a philosophical difference between "being" and "reporting". AI involves a lot of debate about philosophy so expect that. Learning, by one definition, involves a permanent change in behavior. This has to be reported in some way. Since programs are data in lisp this is something of a semantics debate. Is the program "really changed" or just "reporting"? Consider a more "advanced" kind of learning where we use genetic programs to evolve behavior. Clearly you can do this all using data but it is a bit more elegant if you can take "genes" (i.e. slices of code), do crossovers (i.e. merge the slices of code into other slices), and get a new mutated set of "genes". These can be embedded in chromosomes which are just larger pieces of code. Real cells don't use a "data scratchpad", they self-modify. The resulting self-modified lisp code has execution semantics defined by the language standard (well, CL has a standard). The "data representation" does not have standard semantics. Data has semantics relative to the "master smart program thing" you wrote. It turns out that self-modifying lisp code is both interesting and elegant as the "data" is "itself", not some reported-thing. There is a reason lisp dominated the AI world for so long. Many centuries ago I authored a language called KROPS (see [1] below). It was a program that allowed you to represent knowledge using both subsumption (KREP-style) and rules (OPS5-style). Adding a new piece of knowledge caused it to self-modify in a way that allowed both execution semantics and "data" semantics. KROPS is way too complex to explain here but it was very elegant. IBM built a financial and marketing expert system in KROPS on symbolics machines. The point of all of this self-trumpet noise is that I don't believe I could have had the insight to build it in a "data" model. I just let it build itself to evolve a problem solution. There was no data, only program. This isn't intended to be a debate about WHY we might want a self-modifying program. It is a question of whether Clojure, as a Lisp, is sufficiently well-crafted to allow it to self-modify. It has been done in other lisps but I don't yet know how to do it in Clojure. If you only want a "master smart program thing that reports data" that's perfectly fine. You could write that in any language. Tim Daly d...@axiom-developer.org [Daly, Kastner, Mays "Integrating rules and inheritance networks in a knowledge-based financial and marketing consultation system" HICSS 1988, pp495-500] -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
