> -----Original Message----- > From: Matt Mahoney via AGI [mailto:[email protected]] > > On Mon, Nov 24, 2014 at 2:15 PM, Jim Bromer <[email protected]> > wrote: > > An algorithm that could 'predict' bits in arbitrary sequences which it > > had never been exposed to can hardly be called a 'learning algorithm'. > > Yes it is. Are you familiar with AIXI? >
Keep in mind that we exist within a real thermodynamic universe where a program of sufficient complexity needs to economically manage the decision of whether or not bits will be flipped in the act of prediction itself since there is an expense of energy incurred. Just comparing non-weighted k-complexity strings in an energy homogenous non-cyclicly influenced, non-simultaneous, non-coexistent environment without any randomness is really an attempt to pidgeonhole creation of AI into the dualism of universal absolutes. "Universal" is different from "General" in the case of intelligence. General intelligence, AGI, is about efficiency in a thermodynamic world. Universal intelligence is more from a dualist perspective.... IMO. John ------------------------------------------- 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
