Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!seismo!rutgers!princeton!mind!harnad From: harnad@mind.UUCP (Stevan Harnad) Newsgroups: comp.ai,comp.cog-eng Subject: Re: Searle, Turing, Symbols, Categories Message-ID: <214@mind.UUCP> Date: Wed, 12-Nov-86 17:01:22 EST Article-I.D.: mind.214 Posted: Wed Nov 12 17:01:22 1986 Date-Received: Wed, 12-Nov-86 22:13:05 EST References: <158@mind.UUCP> <150@cwrudg.UUCP> <160@mind.UUCP> <2495@utai.UUCP> <1817@rlvd.UUCP> Organization: Cognitive Science, Princeton University Lines: 128 Summary: On (1) whether having a mind is a necessary or merely a sufficient condition for having intelligence; and on (2) the open-endedness of the formal and informal Total Turing Test (TTT) Xref: mnetor comp.ai:9 comp.cog-eng:3 kgd@rlvd.UUCP (Keith Dancey) of Rutherford Appleton Laboratories, Atlas Buildings, U.K. writes: > I don't think it wise to try and equate 'mind' and 'intelligence'. > A 'mind' is an absolute thing, but 'intelligence' is relative. I'm not quite sure what you mean, but perhaps it is that intelligence seems to be something you can have to varying degrees, whereas a mind seems to be an all-or-none phenomenon. If so, I agree. But it's also true that the only handle we have on what might distinguish "mere" clever performance tricks from "true" intelligence is that the latter, at least, is indeed exhibited by humans, i.e., creatures with minds. So the first-order problem is that of distinguishing pseudo-intelligent performance from intelligent performance (if there is any difference), and one "sure" case of the latter is our own. The reasoning then runs like this: (i) I "know" I'm intelligent, because I know what it's "like," first-hand. (ii) I infer that other people have minds like my own, and hence that they too are intelligent; the basis for my inference is that they act the way I do in all respects that seem intuitively relevant (they don't have to be my mirror image, just indistinguishable from me in all the intuitively compelling and rationally justifiable respects, i.e., total performance capacity). (iii) Why deny the same benefit of the doubt to other organisms and devices, if indistinguishable in the SAME respects. (iv) Having ascertained that turing-indistinguishable performance capacity is the basis for my inference about mind, the inference about "intelligence" inherits it. The only refinement I've put on this is the notion of the Total Turing Test -- something that captures ALL of our performance capacity. This is to rule out toy problems, toy models and toy modules, which mimic subtotal fragments of our performance capacity, and leave open a larger-than-necessary possibility that they do so in a significantly different way, namely, unintelligently. To put it another way, having a mind seems to be a sufficient condition for displaying intelligent performance. If it is not also a necessary condition, then our theories have, as I've argued, yet another order of underdetermination in cognitive science, over and above the underdetermination of ordinary scientific theory. Note that even in human beings (and other organisms) HOW intelligent they are is a matter of degree, but THAT they are intelligent at all seems to be an all-or-none accompaniment of being human beings (or other organisms). [Please, for standard objections about mental retardation, aphasia, coma, brain death, etc., and their rebuttals, see prior iterations of this discussion.] > most people would, I believe, accept that a monkey has a > 'mind'. However, they would not necessarily so easily accept that a > monkey has 'performance capacity that is indistinguishable from human > performance capacity'. On the other hand, many people would accept > that certain robotic processes had 'intelligence', but would be very > reluctant to attribute them with 'minds'. I think there is something > organic about 'minds', but 'intelligence' can be codified, within > limits, of course. I agree that monkeys and other organisms have minds. I think they also have intelligence. In prior iterations I suggested that nonhuman variants of the Total Turing Test (TTT) will probably be needed too. These will still have to be "total" though, within the ecology in question. Unfortunately, because not all of us are good naturalists, and because all of us have weaker intuitions for minds in other species than our own, these nonhuman variants will be both functionally more difficult to attain (because of difficulties in knowing the species'total performance ecology) and intuitively less compelling. They may be important way stations on the path to the human TTT Utopia, though. The second part of your statement, about accepting (a priori?) that certain robotic processes have intelligence (and that intelligence is codifiable [i.e. symbolic?]) unfortunately begs the question, which is whether there is in fact some important natural or functional kind of which human intelligence and any old fragment of clever robotic performance can both count a priori as instances. [Since both this atheoretical view you mention and the TTT view I advocate happen to share an ultimate reliance on performance -- clever in one case, turing-indistinguishable from mindful performance in the other -- and since the mind really only plays a shadowy intuitive role in the TTT view, I'm inclined to think that nonmodularity (i.e., the insistence on TOTAL performance capacity) is really the only thing that separates the two views.] --- I will close with a reply to an earlier comment by eugene@aurora.UUCP (Eugene miya) of the NASA Ames Research Center, Mt. View, Ca., who wrote: > No single question can answer the question of intelligence, then how > many? I hope a finite, preferably small, or at least a countable number. The Total Turing Test clearly has to be open-ended, just the way it is when we use it informally in our ongoing provisional solutions to the "other minds" problem. Being only an informal criterion for capturing our intuitions about having a mind (the way Church's Thesis tries to capture our intuitions about "effective procedures"), success on the turing test, even after a lifetime of trials, can be no guarantor of anything. And that's without mentioning the normal inductive risk that forever attends any empirical hypothesis as long as time goes on... The same is true of what I called the formal component of the Total turing Test, i.e., the empirical burden of designing a device that displays ALL of our performance capacities. Here, though, the only liability is inductive risk, which, as I've argued, is just the normal underdetermination of scientific theory. The formal component makes no mention of capturing mind, only total performance capacity. To a good enough approximation I suppose the number of performance tasks the model must prove capable of handling here is finite, though it's probably very large. > [The turing test] should be timed as well as checked for accuracy... > Turing would want a degree of humor... > check for `personal values,' `compassion,'... > should have a degree of dynamic problem solving... > a whole body of psychometric literature which Turing did not consult. I think that these details are premature and arbitrary. We all know (well enough) what people can DO: They can discriminate, categorize, manipulate, identify and describe objects and events in the world, and they can respond appropriately to such descriptions. Now let's get devices to (1) do it all (formal component) and then let's see whether (2) there's anything that we can detect informally that distinguishes these devices from other people we judge to have minds BY EXACTLY THE SAME CRITERIA (namely, total performance capacity). If not, they are turing-indistinguishable and we have no non-arbitrary basis for singling them out as not having minds. -- Stevan Harnad (609) - 921 7771 {allegra, bellcore, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet