Wednesday, November 08, 1995

Learned behaviors and evolution

In response to a note remarking that learned behaviors (such as acquired musical skills or maladaptive stuttering behaviors) could be passed on by evolution.

I haven't read "The beak of the Finch", but I do know that Darwinian evolution
does NOT allow for such a mechanism. This was precisely the debate between
the Darwinians and the Lamarkians (sp?). The former explaining evolution via
chance occurrences the latter as guided by environmental "drive" or "need".
A mechanism for the former was found (the genetic system) but not for the
latter (still, I believe). But this does NOT mean that it is conceptually
impossible. We are, after all, the result of evolution, and we are now doing
precisely that, when we do genetic engineering. The point is that we are now
able to figure out, at least in a few cases, what genetic changes are required
in order to produce desired results. It is very difficult to imagine how
nature could do that simply at the level of the body i.e. without the
social/scientific evolution that allows for our ability to build "models" of the genetic system.

Computer "do" that only in the sense that the programmer knows (or should
know) how to change the code in order to obtain desired effects, but even
they often find it hard to do (ever heard of bugs?). In terms of neural nets
or artificial systems that learn and evolve without conventional programmer's
help, the ability to combine both kinds of evolution (chance and goal direction)
is an open area of research (one I am involved in, by the way).

No comments: