From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!csd.unb.ca!morgan.ucs.mun.ca!nstn.ns.ca!aunro!ukma!darwin.sura.net!europa.asd.contel.com!uunet!mcsun!uknet!edcastle!aiai!jeff Thu Feb 20 15:21:56 EST 1992
Article 3846 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca comp.ai.philosophy:3846 sci.philosophy.tech:2160
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!csd.unb.ca!morgan.ucs.mun.ca!nstn.ns.ca!aunro!ukma!darwin.sura.net!europa.asd.contel.com!uunet!mcsun!uknet!edcastle!aiai!jeff
>From: jeff@aiai.ed.ac.uk (Jeff Dalton)
Newsgroups: comp.ai.philosophy,sci.philosophy.tech
Subject: Re: Implementation (was: Re: Causes and Reasons)
Message-ID: <6203@skye.ed.ac.uk>
Date: 18 Feb 92 22:36:18 GMT
References: <6538@pkmab.se> <6183@skye.ed.ac.uk> <1992Feb16.223809.19674@bronze.ucs.indiana.edu>
Sender: news@aiai.ed.ac.uk
Organization: AIAI, University of Edinburgh, Scotland
Lines: 45

In article <1992Feb16.223809.19674@bronze.ucs.indiana.edu> chalmers@bronze.ucs.indiana.edu (David Chalmers) writes:
>In article <6183@skye.ed.ac.uk> jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
>
>>What I want to know is why does David Chalmers need his restricted
>>notion of implementation?
>>
>>1. It limits his conclusions.
>>
>>For instance, he doesn't conclude that if we write an understanding
>>program in Lisp, then running it on any machine that will run it
>>produces understanding, and if we translate it into Prolog it still
>>works.
>
>I answered this already: A limited conclusion is plenty strong enough
>for the purposes of establishing a strong "strong AI".  I think that a
>less limited conclusion is also true, but arguing for that view is
>more complex and isn't necessary e.g. to argue against Searle.

What I don't understand is why the argument has to go in a way
that makes the more general case so much harder.  (See below.)

And what you seem to end up with is more or less an argument that
brain simulations (at the functional) level work.  This may be
comforting to connectionists and the like, but it doesn't show that
even their programs would suffice if translated into a different
language.

>>2. He doesn't need it to say that programs specify causal systems.
>>
>>He could do that with a broader / weaker notion of interpretation.
>>However, the class of causal systems would be larger.
>
>I agree, I could do it.  It would just take more work, and I don't
>think it's necessary.  Note, however, that the definition couldn't be
>weakened all the way to allow any compiler with I/O equivalence.

But why would it be some much harder?  If all you need is to say
programs are implemented as causal structures, you can get by with a
much weaker notion.  I don't know what in your "crumbly cake"
arguments requires the very narrow notion of implementation.

Now, maybe your arguments with Zeleny require it, but not the
ones against Searle.

-- jd


