Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84; site ccice2.UUCP Path: utzoo!watmath!clyde!bonnie!akgua!sdcsvax!sdcrdcf!hplabs!hao!seismo!rochester!ritcv!ccice5!ccice2!bwm From: bwm@ccice2.UUCP (Brad Miller) Newsgroups: net.ai Subject: Re: AIList Digest V2 #173 Message-ID: <523@ccice2.UUCP> Date: Mon, 10-Dec-84 13:25:24 EST Article-I.D.: ccice2.523 Posted: Mon Dec 10 13:25:24 1984 Date-Received: Thu, 13-Dec-84 01:41:58 EST References: <3708@ucbvax.ARPA> Organization: CCI Central Engineering, Rochester, NY Lines: 27 > Date: Thu, 6 Dec 1984 01:59 EST > From: MINSKY%MIT-OZ@MIT-MC.ARPA > Subject: Infant Amnesia V2 #165 > > ... I like theories like this: our experience is first > encoded in rather stupid ways; a square is seen as a line attached to > another line attached to another line, etc. Like an early > assembly-language. Later, a square is represented as "closed path of > equal lines" and, later, orthognal pairs of parallels, etc. -- going > to Fortrams to Pascals to LOGs to SMALLTalks to who-knows-what. The > representatins and their interpreters grow more sophisticated, and > those first machine-languages of infancy just can't be always > upwards-compatible. So, even if those early memories were not, in > fact, entirely ever lost, they're doomed to become > unintelligible, eventually. This doesn't explain how someone with an edictic or photographic memory can examine a scene after being exposed to it and discover things about it. The theory I like says that memories of scenes or situations are stored as holograms. We simply garbage-collect our earlier memories due to lack of access, that is, we haven't stored them as instances of anything, so there are no pointers to them. Brad Miller -- ...[rochester, cbrma, rlgvax, ritcv]!ccice5!ccice2!bwm