Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!asuvax!ncar!ames!amdahl!kp From: kp@uts.amdahl.com (Ken Presting) Newsgroups: comp.ai Subject: Re: "self" consciousness Summary: Correctness of self-description is judged as in everyday life Keywords: Turing test, consciousness Message-ID: <40ak02xa865Y01@amdahl.uts.amdahl.com> Date: 6 Feb 90 19:19:48 GMT References: <15439@well.UUCP> <11673@csli.Stanford.EDU> <11324@venera.isi.edu> <1700@castle.ed.ac.uk> <11489@venera.UUCP> <6340@sdcc6.ucsd.edu> <11849@csli.Stanford.EDU> <6371@sdcc6.ucsd.edu> <93Nb02wq81R.01@amdahl.uts.amdahl.com> <1758@oolong.la.locus.com> <25a702Kz Reply-To: kp@amdahl.uts.amdahl.com (Ken Presting) Distribution: comp Organization: Amdahl Corporation, Sunnyvale CA Lines: 65 (I took out the smileys, but it's nothing personal) In article <3553@uceng.UC.EDU> mfinegan@uceng.UC.EDU (michael k finegan) writes: >>kp@amdahl.uts.amdahl.com (Ken Presting) writes: >>Let me restate the self-description scale: >> >>For any object, there is a set of sentences which refer to that object. >>For any two objects, if object A can correctly affirm or deny each >>sentence which refers to itself when object B can do the same (mutatis >>mutandis), and object A can correctly affirm or deny some sentences which >>object B cannot, then A has superior self-descriptive capacity than B. > >Seriously, this sounds fine, but who judges 'correctness' ? A very intelligent >source (machine or otherwise) might be correct, with the judge incapable of >realizing it. Examples: a prodigous student and their teacher, Gallileo and >the church, etc. > >>It's not formally a part of the criterion, but I should say that ordinary >>physical description of one's body and its relation to the surroundings >>is the main thing I have in mind. I should also note that this is *not* >>the only criterion relevant to judgements of intelligence. The last paragraph is the answer to the question. If the self-descriptive statements are about mundane facts (like "My hair is brown" or "My cabinet is blue") then there should be little controversy over the right answers. Most authors seem to favor asking deep subjective questions in the Turing Test, but that seems unproductive to me. ELIZA gave a good impression of depth, at least to some people. A good example a a computer describing itself accurately is the Unix(tm) "ps" (process status) command. At least the machine knows what it's doing! Process accounting statistics are even better; they involve memory of past actions. I'm using this example to show that self-description is only part of the story for a criterion of intelligence. The question of who is a reliable judge for Turing-type tests has been around awhile. I've always liked the suggestion that programmers would be better able to detect programs than psychologists. Realistically, I'll bet that nobody will trust anybody's judgement except their own, and the wrangling will get nastier as programs get smarter >> Whether or not this scale has any relation to the various notions of >> consciousness proposed by psychologists and philosophers is a separate >> (but not unimportant) question. >Why do you guys always ignore the biological notions ? I'm not sure what you mean. Most biologists have better things to do than debate the nature of consciousness, like bar-coding bees, or stealing food from _Sphex_. >> I do think this scale agrees roughly with everyday talk about >> consciousness - when you're asleep (ie unconscious) you don't have much >> to say about yourself (or anything else). Rocks have nothing to say, >> chimpanzees have only a little. >The concept of inner dialogue has been around for a while (cf. psycho-analysis). >This might be a necessarry pre-requisite for a computer to pass the Turing test >(namely, create an inner model of the enviroment, try out its planned responses >within this model, and see what happens - then actual present the critiqued >response to the enviroment). This is in line with a.i. developments in knowledge >representation. This sounds like a useful technique to me. It resembles gedanken- experiments, design reviews, and everyday planning of tasks. An inner dialog in itself is not what I mean by self-description, though. A system could use a fancy inner-dialog algorithm and still be mostly oblivious.