Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!helios!bcm!dimacs.rutgers.edu!rutgers!rochester!uhura.cc.rochester.edu!troi.cc.rochester.edu!ta2cs220 From: ta2cs220@troi.cc.rochester.edu (H. Y. Firooznia) Newsgroups: comp.ai.philosophy Subject: Re: computer life? Message-ID: <12616@ur-cc.UUCP> Date: 5 Mar 91 22:28:53 GMT References: <12565@ur-cc.UUCP> <1991Mar4.143106.8838@msuinfo.cl.msu.edu> <1991Mar4.144332.9415@msuinfo.cl.msu.edu> Sender: news@uhura.cc.rochester.edu Organization: University of Rochester, Rochester NY Lines: 112 In article <1991Mar4.143106.8838@msuinfo.cl.msu.edu> dailey@galaxy.cps.msu.edu (Chris Dailey) writes: >In article <12565@ur-cc.UUCP> ta2cs220@troi.cc.rochester.edu (H. Y. Firooznia) writes: >>In article <1991Mar3.025707.16737@msuinfo.cl.msu.edu> dailey@galaxy.cps.msu.edu (Chris Dailey) writes: >>>In article <12548@ur-cc.UUCP> ta2cs220@troi.cc.rochester.edu (H. Y. Firooznia) writes: >> ... >>issue of whether or not it exhibits intelligence. You seem to be defining >>intelligence on the basis of internal operations, as opposed to behavior. > >So is a large database program intelligent? I'd say not. It just has >the mechanics to calculate and then access a lot of information. It >almost never makes inferences to find out more information unless the >programmer specifically tells it to do so (including telling it HOW to >do so). My definition for intelligence certainly goes beyond being able >to do direct calculations from easy-to-access data. Okay. Again, the issue here seems to be one of behavior vs. internal states. I don't know if the database program is intelligent. If, however, it was capable of conducting a conversation, I might tend to say it is intelligent. What if the program displayed intelligent behavior? Then what? [text deleted] >intelligence? For intelligence one of the questions I asked was about >whether a plant that moves to tilt toward the sun during the day has >intelligence. I am not sure, because I do not know how and why a plant >does this. I would definitely say the plant shows intelligence, but I >do not know if it actually HAS intelligence. [These two concepts are >not mutually exclusive in my mind.] When you say "has intelligence", you seem to be implying some specific form of mental operation, as opposed to behavior. If your definition has such requirements, then I suppose a plant, by that definition, would not be intelligent. I'm curious. What do you think of the Turing Test? It seems, from your definition, that a computer passing the test is not necessarily intelligent. >I would say that to actually have intelligence (and not JUST DISPLAY >intelligence) a chess program would have to be able to learn... to go >beyond the known and into the unknown... to evaluate its performance >and be able to improve on that. I think that's already been done. As to learning, and in particular discovering the "unknown", some chess programs can do that too. Excerpted from "The world's next chess champion?", Popular Science, 3/91: " Even strong players can learn from computers. Karpov used a machine, hand built by Fidelity Electronics International and equipped with Motorola's new 68040 microprocessor, to prepare for the championship he narrowly lost to Kasparov in January. 'I started in this believing that humans understood everything about chess,' says Carnegie-Mellon computer scientist Hans J. Berliner, a former World Correspondence Chess Champion and the designer of Hitech. 'I was wrong. Machines can find things humans haven't even dreamt about.'..." ... >>What if a computer, >>given no instruction on how to play chess, ended up learning to play >>as Deep Thought does? What's the difference? > >Could it, though? It would have to come up with some strategy that >would be [perhaps only roughly] equivalent. The difference is NOT >in the way the computer/program plays chess. The difference is in >WHERE THE STRATEGIES COME FROM! Okay. So in a nutshell, you're saying: For something to be intelligent, it has to learn its strategies itself. Correct? >We believe that something is intelligence because it displays >intelligence. If something displays intelligence, we often assume that >it has intelligence because it learned how to do that. Isn't that why >the Eliza and Doctor programs are so well known? Because they could >fool people at least a LITTLE bit into believing that the computer had >actually learned how to converse with you? Then when you learn the >tricks used to create the program, you can then convince yourself that >it was not intelligent. I would convince myself that the Eliza program was not intelligent in the way that I had originally perceived it to be intelligent (human way). As to whether an Eliza program is actually intelligent, I guess that depends upon your definition. Basically, I'm saying that there is no well-defined concrete thing as intelligence- it's just a general property that we observe in some living things, humans in particular. It seems that, as a result, our definition of intelligence is generally confined to human-like intelligence. So, if you're saying that Deep Thought, or Eliza, or whatever, isn't intelligent in that it isn't doing what humans are doing internally, then it isn't intelligent. But, what do you say in cases in which things aren't so well defined? -e.g. What if someone could, instead of learning things directly, have the knowledge gained by others directly implanted in their head? Would they then _not_ be intelligent? Perhaps intelligence is not so much the learned strategies, as it is the _ability_ to learn new strategies? Is this the difference between computer chess players and humans that you were talking about? If so, then what if a computer could learn new strategies on its own? Would it then be intelligent? I think the problem here is, again, when we speak of "intelligence", we really mean "human intelligence". -Hoss