On 2013-10-09 14:48, Matt Mahoney wrote:
https://docs.google.com/drawings/d/1ZaTlA92vUwCzKFXh_wOnFAVrGLGgzUyDj3wNfUS-PWI
I think Ai is "unsolved" because it's badly framed. So a lot of it seems oriented around mid 20th century behaviorism or rational utility maximising agents. Those things may be interesting, but they're off-base if you want to produce something similar to human intelligence.
From a mid 20th century perspective much of human behavior seems baffling and hopelessly irrational. To produce a human-like cognitive system I think it needs to include more social and narrative elements, such as the maintenance of a social graph (Dunbar, etc) and a language system capable of supporting internal dialogues and imaginative narrative reconstructions - what I call "circumstance based reasoning". Instead of trying to maximize utility probably what people are mostly doing is trying to maximize the consistency or desirability of their autobiographical narrative relative to other known historical or fictional accounts, or simply to live according to some particular principle or aesthetic (both of which also have a historical basis to them). Outside of being embedded in a multi-agent system which has a history it's difficult for that sort of cognitive system to appear.
------------------------------------------- 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
