Path: utzoo!attcan!uunet!mcvax!ukc!etive!aiai!jeff From: jeff@aiai.ed.ac.uk (Jeff Dalton) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Message-ID: <221@skye.ed.ac.uk> Date: 28 Feb 89 20:04:21 GMT References: <4298@pt.cs.cmu.edu> <51123@yale-celray.yale.UUCP> <563@aipna.ed.ac.uk> Reply-To: jeff@aiai.UUCP (Jeff Dalton) Organization: AIAI, University of Edinburgh, Scotland Lines: 55 In article <563@aipna.ed.ac.uk> rjc@uk.ac.ed.aipna (Richard Caley) writes: >So it _is_ a definitional problem. Since we have assumed that the >behaviour is identical whether or not it understands, we must rely on >deduction based on the structure of the system to tell us if it >understands. Most significantly, we can't rely on the method we use for >humans - if we ask the room ( presumably in chinese ), it says yes, >otherwise the behaviour is not like that of a native speaker! You haven't said anything that shows it's a definitional problem. I do not see how much can be gained by moving from "does X understand?" to "what does 'understand' mean?" In particular, answers to the second question will not necessarily let us resolve the first. It is always open for someone to say "well, your definition of 'understand' is wrong because there's a counterexample: X isn't doing something that satisfies your definition, but X is understanding." And then we're right back where we started. Or, to look at it another way, what does it matter whether something is called "understanding" or not? What really matters is whether things are the same or different in some interesting way. Think of yourself. If you understand English and do not understand Chinese, do you accept that there's some difference there? Do you have to define "understand" before you can answer, or do you already have an adequate understanding of "understand"? >Without defining understanding we can't argue with it since our >intuitive knowledge of understanding is only for _ourselves_, we apply >it to other people since they seem rater similar, we can _try_ >and apply it to philosophers in rooms or computer systems but I would >not trust the result - Suppose we had the sort of definition you want, and suppose it let us say that X understood and Y didn't. Then someone might say "well, I guess 'understanding' wasn't the right thing to ask about after all." The reason we have problems deciding about computers and philosophers in rooms is that we don't know all that much about how our minds work and because we can never get inside someone else's subjective experience. Definitions of "understanding" do not help with either problem. >Aren't you assuming the result here. If searle running the program is >"-" WRT "understanding" the naturally the system does not understand. You're right here. You're also right that the structure of the system, or something like that, may turn out to be significant. But I think that's about all we can say at this point. After all, we don't have any machines that behave as if able to understand Chinese, so it's hard to say anything about their structural properties. -- Jeff