Decius wrote:
] ] When the internal architecture of a trained
] ] artificial neural network is gradually relaxed or
] ] destroyed, that network tends to spontaneously
] ] produce a succession of "impressions" from it's
] ] learned knowledge domain. I refer to this state
] ] as "dreaming." A dreaming net's output stream
] ] often holds a mixture of both straightforward and
] ] hybridized exemplars from its training set. If we
] ] allow a second neural network to watch for any
] ] useful concepts that emerge from the first, we
] ] form a so-called "Creativity Machine." Creativity
] ] machines may perform remarkable feats of
] ] invention and discovery, ranging from the composition of
] ] music to the prediction of totally new ultrahard
] ] materials.

well, here is my analysis...he's basically re-invented genetic programming...but he's doing it in a slightly novel way...I've seen similar things to his where genetic algos are using neural nets as their fitness function, and to be honest, if his noise is random then he will eventually get to a solution, but most GAs have a way of "evolving" the noise to get a faster solution...

at best id say this is a new form of evolutionary programming...

that isnt to say it isnt cool, its really cool, and he's probably done more in the area than most of the people ive read about, but i see nothing to lead me to believe this isnt simply an evolving neural net, and thus yet another form of evolutionary programming...

that said i do see some clear limitations in his design...namely, if my underderstanding of what he's doing is correct, the NN will simply be able to come up with dirivative works based on its training...this is very useful indeed, and in some ways will get him to many improvements on existing knowledge, but I think it lacks the ability to come up with original "ideas"...dirivative works are important, but other AI models like genetic algos have the ability to come up with radically new solutions to problems that are not based on existing designs or knowledge...

a short while back i read about someone doing work on genetic algos using NNs as the fitness function, I'll dig back into my history and meme it when i get home...

--Abaddon

RE: A technical discussion of Creativity Machines