News for hackers. Stuff that matters.
Re:What intelligent machines need to incorporate from the ne... by mindcrime on Friday June 02, 2017 @11:37PM (#14474894)
True. But, conversely, Hinton himself says in some of his videos (sorry, don't have the specific link in front of me) something to the effect of "ANN's are only very loosely modeled on real neurons and we make no pretense that this is anything like the way a real brain works". (Paraphrased, possibly poorly).

The sense I get is that Machine Learning is mostly an empirical field at the moment, without a terribly solid theoretical underpinning. Which is not, of course, to suggest that ML researchers haven't consulted neuroscience at all. But there still seems to be a pretty big disconnect between neuroscience and ML. Unless I've just really missed something, which is entirely possible.

Re:What intelligent machines need to incorporate from the ne... by closed on Saturday June 03, 2017 @12:34AM (#14475026)
Yeah, that's fair. The goals of machine learning and neuroscience seem fairly different. Nobody in ML will complain if your neural network has great prediction but isn't supported biologically. OTOH, neuroscientists might take a model that isn't the best in terms of prediction, if it seems to better represents how the brain operates.

There seems to be some interesting interplay, though. For example deepmind recently sponsored this ANN conference where most the speakers were neuroscientists.

https://sites.google.com/site/ncpw15/