Path: utzoo!attcan!uunet!lll-winken!csd4.milw.wisc.edu!nic.MR.NET!umn-d-ub!rutgers!elbereth.rutgers.edu!harnad From: harnad@elbereth.rutgers.edu (Stevan Harnad) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Summary: On Human Rights for Scratches On Paper Message-ID: Date: 1 Mar 89 18:44:45 GMT References: <4298@pt.cs.cmu.edu> <18073@iuvax.cs.indiana.edu> Organization: Rutgers Univ., New Brunswick, N.J. Lines: 81 dave@cogsci.indiana.edu (David Chalmers) of Concepts and Cognition, Indiana University wrote: " [Concerning symbolic modeling vs neural modeling] I didn't think this " was the time or the place for a switch of topic to an issue far more " complex than Searle's misleading intuition pump. No switch. That IS Searle's topic. It is most respondents who have over-simplified it. " Do you believe neurons (taken alone) have semantics[?] [I take it the " answer has to be "No."] Do you believe the brain as a whole has " semantics[?] [I take it the answer is "Yes."] Given this, you must " accept that semantics can arise out of non-semantic objects... " not so much those objects as the complex patterns that they form. Of course semantics arises out of nonsemantic objects. But there are nonsemantic objects and nonsemantic objects -- and scratches on paper (even when implemented as symbol-crunching computer programs) do not seem to be the right kinds of objects. Likewise there are patterns and patterns. My "Robotic Functionalism" IS a form of functionalism -- it does hold that cognitive function is some "pattern" of physical function. But, unlike standard "Symbolic Functionalism," it denies that that pattern of physical function consists merely of formal symbol manipulation. It can be SIMULATED by symbol manipulation; but if what is simulated is not merely symbolic function (e.g., if an essential part is analog processing) then it cannot be IMPLEMENTED as just symbol manipulation. (And, as I said in my postings and article: Only implemented planes/brains can fly/understand.) " The fact that complexity is a necessary condition for information would " suggest that appeals to complexity are not mere hand-waving. But necessary conditions are not sufficient conditions. And mere complexity will not you a mind get. There's complexity and complexity; and a lot more conceptual work to do before you have a viable model for the mind. " [Harnad's "Robotic Functionalist Reply"] strikes me as rather like the " point-missing "Robot Reply" in Searle, despite your disclaimers. I " thought that the "Stephen Hawking argument" was a rather good reply to " this stuff. What's important for subjective experience is a brain " state, not a bodily state; and AI claims to be able to simulate any " brain state whatsoever According to Robotic Functionalism, the device -- the "inner core," the "brain-in-a-vat," or whatever you like -- that will be able to successfully pass the Linguistic version of the Turing Test (LTT) (symbols-in, symbols-out) will have to have and draw upon the internal causal wherewithal to pass the Total (robotic) Turing Test as well (even if it does not have to display it behaviorally). I'm sure Stephen Hawking has that inner core; and it's just a current blinkered fantasy that that inner core consists of nothing but a symbol-cruncher! Hawking's intact internal nonsymbolic (brain) functions are crucial to his having a mind whether or not he can or does display them in any other form than a verbal one. (Or didn't people in AI know that if you yanked off from the brain the "body" and all its sense organs -- some of which happen to be PART of the brain, by the way -- you weren't just left with a digital computer?) " if we were all made of paper we'd say the same thing: "It's easier and " safer to assume that neuro-thingies don't support TRUE experience; and " after all we have no direct evidence for it, only their meaningless " claims. So lets just ignore anything which these systems have in " common (viz. extreme complexity, intelligent behaviour) and just " concentrate on their differences."... I don't want to be inflammatory, " but it sounds not unlike many an argument used by a racist in days gone by. And if my grandmother had wheels, or the world were one-dimensional, or stones had minds... You can't make a counterfactual and implausible conclusion seem more plausible by simply adopting it as a premise. Ref: Harnad (1989) Minds, Machines and Searle. Journal of Experimental and Theoretical Artificial Intelligence" 1: 5-25 -- Stevan Harnad INTERNET: harnad@confidence.princeton.edu harnad@princeton.edu srh@flash.bellcore.com harnad@elbereth.rutgers.edu harnad@princeton.uucp BITNET: harnad@pucc.bitnet CSNET: harnad%princeton.edu@relay.cs.net (609)-921-7771