Path: utzoo!utgpu!utstat!jarvis.csri.toronto.edu!mailrus!uflorida!haven!aplcen!jhunix!ins_atge From: ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Summary: Homoncleus in the Neuron Keywords: Homoncleus, Comprehension, Learning Message-ID: <1082@jhunix.HCF.JHU.EDU> Date: 10 Mar 89 05:02:30 GMT References: <4298@pt.cs.cmu.edu> <9787@ihlpb.ATT.COM> Reply-To: ins_atge@jhunix.UUCP (Thomas G Edwards) Organization: The Johns Hopkins University - HCF Lines: 56 In article harnad@elbereth.rutgers.edu (Stevan Harnad) writes: >arm@ihlpb.ATT.COM (Macalalad) of AT&T Bell Laboratories wrote: >" the only point you didn't comment on was the argument that neurons seem >" to be the "chalk-pushers" of the brain, yet they individually don't >" seem to have understanding... Instead of having just one person >" in the Chinese room, let's have a lot of people, comparable to the >" number of neurons in the brain. All of them are busy doing calculations >" on the chalkboard and passing pieces of paper around, all in strict >" adherence to their rulebooks. >To me there is only one thing that sounds more unlikely than the notion >that the LTT could be passed by symbol-crunching alone, and that is the >notion that all neurons do is crunch symbols. But that's certainly one >way of supporting the proposition that symbol-crunchers can understand. >It's called argument by assumption (i.e., it's the same circular >reasoning we keep encountering over and over on this topic). >For the record, the only reason the real brain is immune to Searle's >Argument is that neurons are NOT just "chalk-pushers." Upon what evidence do you base your statement "neurons are NOT just 'chalk pushers.'"? I'll be the first to admit that neurons do have a complicated transfer function between input and output, but your statement gives them a little more cognitive ability than I'm ready to accept. Creatures with relatively few neurons show little cognitive ability (Aplysia for instance). Or perhaps our definitions of chalk-pushers is a little different. I assume that by "chalk-pusher" you mean someone with memory and a releatively simple transfer function between input and output (perhaps a little more complicated than 1/(1+e^(-x)), but nethertheless much less complicated than that of any relatively lucid [in terms of being able to deal with real world problems] brain, including Aplysia, grasshoppers, etc.) From my desperate attempts to find neural net literature five years ago, I know there are hoards of books desribing mathematical models of real neurons (probably not including neuropharmacology, which _is_ important, but including nerual potential values). Anyway, the main idea here is that neurons are not homonclei!. They cannot do any kind of reasonable cognitive task without being in a weighted network with many, many others. The neruon does not understand, the network _does_ (unless you are ready to drop the concept that people understand). Is this not a good precedent to show that systems do exist in which the parts do not understand, yet the system does? Furthermore, on this ever "angel on the head of a pin" debate, I believe that the subjective concept of understanding is not scientifically provable as per the non-performance oriented concept. If something is in a opaque box, how can we know if it understands or not? It could just as well be a cat, cow, rock, or person. If we think we can determine whether the mind understands or not while we keep it in a box, we are deluding ourselves. A performance test based on input and output which shows real-world reasoning is the best we can ever hope for. -Thomas Edwards ins_atge@jhunix