On Sat, Apr 26, 2008 at 7:28 PM, Mark Waser <[EMAIL PROTECTED]> wrote: > Now, the next argument is -- how effective are these systems, both in > general and at being general? > > TD-Gammon seems to be a pretty limited application (i.e. it's bad at being > general) and spam filtering tries to be general (though, this is arguable > since most spam filters actually use a small bunch of human-generated tricks > to generate a large number of rules) but has an effectiveness level far > below what is desired.
I would say spam filtering is like TD-Gammon: effective enough to be good for its specific purpose, but not very general (it doesn't do anything except filter spam, after all). > Richard's argument is that a system that has all four of his features of > doom is going to be too complex to extend to general effectiveness and > effective generality. Well, he also argued that a system that has all four of his features of doom can't be understood; and I've disproved that much by pointing to such systems that were understood well enough to successfully build them. > I'm not sure that I agree (in fact, I don't) but there aren't any systems > out there that clearly disprove his thesis. Well, that's only because no general intelligence of any kind exists today, other than humans. If that's the level of disproof being sought, then we can't disprove the thesis "a machine can't think because it doesn't have a soul" either. All we can do in such cases is point out that the brain works, so a machine operating in the same universe under the same laws of physics should in principle be able to do similar things. ------------------------------------------- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
