Yes, learning the exact connectivity of the brain would be extremely difficult. But that doesn't mean getting substantially better connectivity information than we have now would not be extremely valuable.
Yesterday I posted the following to this list. It shows IARPA thinks getter more exact information about the brain could be helpful in designing NN architectures. "Interesting article from Kurzweil AI about IARPA funding of project to do detailed study of brain's visual system. Studies like this could make great advances in our ability to understand how to make powerful neural net artificial intelligences. " http://www.kurzweilai.net/cmu-announces-research-project-to-reverse-engineer-brain-algorithms-funded-by-iarpa " On Sun, Feb 14, 2016 at 9:06 PM, TimTyler <[email protected]> wrote: > On 2016-02-14 17:42, EdFromNH . wrote: > > I heard the author of this book on a Dallas/Fort Worth NPR interview show >> called "Think". In it he poo-pooed the importance of learning the >> connectome, saying the connectome wasn't really important, that what was >> important was learning the brain's algorithm. >> >> It is stunningly dumb to minimize the importance of the connection >> architecture of a neural net, particularly if you include connection >> weights in the connectome [...] >> > > It depends a bit on what you are trying to do. Most NN researchers left > biological > realism behind long ago when they adopted back propagation. No doubt there > are > some remaining secrets buried in the connectome - but the issue is the > time and > resources it will cost to find them that way - instead of via reinvention. > The physical > brain is quite a messy tangle; it's not too attractive a place to go > fishing for insights. > > -- > __________ > |im Tyler http://timtyler.org/ > > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/8630185-a57a74e1 > Modify Your Subscription: > https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
