>>14422515you do know the brain has never proven to be binary, right? behaviorists latched onto the action potential as an analogue to computer signals without considering the possibility (and immense evidence for) neurons individually not being deterministic state machines based on incoming external neuronal signals - basically the only neurons that act this way are signal repeaters.
the ANN approach is based on a certain model of the brain, yes, and it's a model that lends itself extremely well to niche problems with the potential for clear and consistent training data, but generalizing it outside of binary "win/lose" contexts is proving to require prohibitively large training sets and network sizes for each new problem
there still hasn't been proof that the output of GAN systems isn't being misconstrued by human judges out of the very well documented human ability to parse patterns out of noise, rather than some indicator of a step towards AGI
