tjf@lanl.gov (Tom J Farish) (01/05/90)
My apologies if this is the wrong group in which to post this question. I recently learned of a set of decision-making processes based on a model of Darwinian 'natural selection'. I would very much appreciate help finding out more about this....references, papers, authors etc. The point (I think) is to use a rule-modifying system in which each generation of rules is derived from the one before in a particularly directed way, based somehow on 'natural selection principles'. Please e-mail to me directly as I do not get a chance to read these groups extensively. A t D h V a A n N k C s E (Thanks in Advance) [Moderators' note: Even though the above request on Darwinian evolution does not refer to objects studied in comp.ai.shells, it directed our attention to the meta question of the evolution of ES shells. What are the selection criteria permitting a shell to survive in the present environment of growing interest in ESs and of hundreds competing shells? Will there be new ecological niches in a changing AI market in which integration with conventional software and neural networks is becoming more important also for ESs? Can we anticipate a decline of the shell market due to overgrowth, unserious company offers, oversized (& priced) dinosaur shells, and new AI fashions such as connectionism? How did evolution proceed from 'AI languages' to 'AI shells', and what term will be the next buzzword? Can we accelerate evolution by introducing a kind of "quality control" for shells, conducting shell contests, speaking more openly about *problems* of shells (what about efficiency, any benchmarks better than "Monkey & Bananas"?)? Please answer to the net both for the original object-level question (the last time we allow non-shell postings) and for these meta-level questions. We will give Tom J Farish a note that he should further watch comp.ai.shells, if only to see what happened with his request.]