afbst@unix.cis.pitt.edu (Audrey F Bancroft) (07/17/90)
Once again, I have a yearning for some nebulous concept. Coupled with my
faith in 'the wisdom of the net', I'll throw this out:
Recently I've been reading 'The Act of Creation' by Koestler, and
have become intrigued with the idea of understanding the moment of
insight. I think it would be fascinating to know more about these
instances of conceptual breakthrough, and particularly to attempt
an integration of the occurence from several areas of Cognitive
Science, such as Neuroscience, Psychology, Philosophy...
Any ideas?
PS -- thanks for the wonderful thoughts on redundancy. I'm compiling a list\
of some and will send it out to those who asked
Thanks in advance
--
**********************************************************
* afbst@unix.cis.pitt.edu *
* "I am a romantic in a world of pragmatists." *
********************************************************** jro@cs.exeter.ac.uk (Jonathan Rowe) (07/18/90)
In article <25960@unix.cis.pitt.edu> afbst@unix.cis.pitt.edu (Audrey F Bancroft) writes: >Once again, I have a yearning for some nebulous concept. Coupled with my >faith in 'the wisdom of the net', I'll throw this out: > > Recently I've been reading 'The Act of Creation' by Koestler, and > have become intrigued with the idea of understanding the moment of > insight. I think it would be fascinating to know more about these > instances of conceptual breakthrough, and particularly to attempt > an integration of the occurence from several areas of Cognitive > Science, such as Neuroscience, Psychology, Philosophy... > > Any ideas? Sure. I'm interested in creativity from an AI / Cog. Sci. approach. One of the main research groups in the field is headed up by Douglas Hofstadter. For some of his ideas, have a look at the following chapters of "Metamagical Themas": Chapter 12 - Variations on a Theme as the Crux of Creativity Chapter 23 - On the Seeming Paradox of Mechanizing Creativity Chapter 24 - Analogies and Roles in Human and Machine Thinking The basic idea is that creativity is about making variations on a standard concept, and using analogies to help work out what variations are possible. The emphasis is on low-level activity producing analogies as emergent phenomena. It isn't quite connectionism, but it's certainly different from classical A.I. I found Koestler's book interesting but insubstantial. The basic idea sounds fine - insight occurs when two seperate "domains" are brought together. But have you thought of just _how_ these domains could be represented, and how they could be brought together (in a computer program, say)? Koestler (I think) believes that the action is controlled (?) by "the ghost in the machine" and so offers no further explanation. Another big problem is how those domains get selected for combination in the first place. Random choices just lead to garbage. Some relevant (but frustratingly inconsistent) AI work is Lenat's programs AM and Eurisko (see, for example, "The nature of heuristics" I, II and III in 'Artificial Intelligence' nos. 19 and 21). Here, the selection and combination of knowledge structures is heuristically guided. However, most of the seemingly impressive results turn out to be a fix. From a cognitive science viewpoint, try "Cognition and Consciousness" by Martindale. This has quite a good chapter on these issues. Jon Rowe. (jro@cs.exeter.ac.uk)