pnettlet@gara.une.oz.au (Philip Nettleton) (06/26/91)
In article <1669@ucl-cs.uucp>, G.Joly@cs.ucl.ac.uk (Gordon Joly) writes: > > Could this thread go into comp.ai.philosophy and die in comp.ai? > > Gordon. Gordon, I agree in principal with the request in relation to most of the mumbo-jumbo that has been discussed on these topic but the original version of 'A "working definition" of intelligence' clearly has its place in both groups (where it was posted). That is unless you believe the "intelligence" in "artificial intelligence" has nothing to do with real intelligence. Perhaps you are even suggesting that a definition of intelligence is not worthy of consideration by researchers in AI? One wonders whether you even took the time to read the definition and form you own opinions, perhaps even adding some more meat to it if you considered it necessary, with an 'appropriate' reply. Philip.
ISSSSM@NUSVM.BITNET (Stephen Smoliar) (06/27/91)
In article <7163@gara.une.oz.au> pnettlet@gara.une.oz.au (Philip Nettleton) writes: > Perhaps you are even suggesting that a definition of intelligence >is not worthy of consideration by researchers in AI? I'm sure Gordon has an aphorism ready for this one; but I shall repeat my own words and reply that a better description would be "a waste of time." There has been nothing which has appeared in the arguments on either comp.ai or comp.ai.philosophy which has supported the case that a definition of intelligence will contribute to our ability to build (artificial) intelligent agents. The question is less appropriate to the arena of scientific argument and more suited to wherever we like to hang out to get a good drink. =============================================================================== Stephen W. Smoliar Institute of Systems Science National University of Singapore Heng Mui Keng Terrace, Kent Ridge SINGAPORE 0511 BITNET: ISSSSM@NUSVM "He was of Lord Essex's opinion, 'rather to go an hundred miles to speak with one wise man, than five miles to see a fair town.'"--Boswell on Johnson
jamesm@hornsby.cs.rpi.edu (Michael James) (06/27/91)
In article <9106270854.AA15777@lilac.berkeley.edu> ISSSSM@NUSVM.BITNET (Stephen Smoliar) writes: >In article <7163@gara.une.oz.au> pnettlet@gara.une.oz.au (Philip Nettleton) >writes: >> Perhaps you are even suggesting that a definition of intelligence >>is not worthy of consideration by researchers in AI? > ... stuff omitted... >has been nothing which has appeared in the arguments on either comp.ai or >comp.ai.philosophy which has supported the case that a definition of >intelligence will contribute to our ability to build (artificial) intelligent >agents... ...other stuff omitted... >Stephen W. Smoliar It seems that the areas in which the AI research community has made the most progress over the past few decades are the ones in which we had EXPLICIT definitions of what was being modelled or simulated. Definitions can vary in their degree of vagueness. Surely if you take a vague definition of intelligence like the Turing Test and then say "Oh, gee, I think I'll go code 'intelligence' in C on my workstation now," you're going to fail. All of the definitions of intelligence I have seen share this same "Intelligence, (wink wink) you know what I mean" attitude. What we need are some very explicit definitions of what we're shooting for. I'm not at all certain that intelligence is explicitly definable, but if its not we should define what it is we want to create and then not really worry about whether anyone calls it intelligent or not. mj ---------------------------------------------------------------------- Mike James Rensselaer Polytechnic Institute jamesm@turing.cs.rpi.edu ----------------------------------------------------------------------
epstein@sunc2.cs.uiuc.edu (Milt Epstein) (06/28/91)
In <nhgl6q_@rpi.edu> jamesm@hornsby.cs.rpi.edu (Michael James) writes: >[some discussion of whether it's worthwhile to define intelligence deleted] > > It seems that the areas in which the AI research community has >made the most progress over the past few decades are the ones in which >we had EXPLICIT definitions of what was being modelled or simulated. >Definitions can vary in their degree of vagueness. Surely if you take >a vague definition of intelligence like the Turing Test and then say >"Oh, gee, I think I'll go code 'intelligence' in C on my workstation >now," you're going to fail. All of the definitions of intelligence I >have seen share this same "Intelligence, (wink wink) you know what I >mean" attitude. > > What we need are some very explicit definitions of what we're >shooting for. I'm not at all certain that intelligence is explicitly >definable, but if its not we should define what it is we want to >create and then not really worry about whether anyone calls it >intelligent or not. I think you may be confusing cause and effect here. What I understand you to be saying is: 1) the AI areas with the most success use explicit definitions 2) therefore, for a better chance of success, we should use explicit definitions 3) therefore, we should come up with an explicit definition of "intelligence" The problem with this is that "intelligence" in general may be too broad of a phenomena to study or model, and an explicit definition of it may not be possible or worthwhile. It just may be that the areas where there has been more success are sufficiently narrowed down that they can use explicit definitions. (In other words, if a range of phenomena is too broad to have an explicit definition, you can't force an explicit definition on them; however, you can narrow down the range of phenomena you're studying so thay they do have an explicit definition.) I apologize if I misinterpreted your statements. >mj >---------------------------------------------------------------------- >Mike James Rensselaer Polytechnic Institute >jamesm@turing.cs.rpi.edu ^^^^^^ (Whoops, perhaps you have it on higher authority :-). -- Milt Epstein Department of Computer Science University of Illinois epstein@cs.uiuc.edu