From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!uunet!mcsun!uknet!edcastle!aiai!jeff Tue May 12 15:49:48 EDT 1992
Article 5495 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!uunet!mcsun!uknet!edcastle!aiai!jeff
>From: jeff@aiai.ed.ac.uk (Jeff Dalton)
Newsgroups: comp.ai.philosophy
Subject: Re: Systems Reply I (repost perhaps)
Keywords: AI Searle Dickhead Barf
Message-ID: <6689@skye.ed.ac.uk>
Date: 8 May 92 21:09:09 GMT
References: <523@tdatirv.UUCP> <6638@skye.ed.ac.uk> <3@tdatirv.UUCP>
Sender: news@aiai.ed.ac.uk
Organization: AIAI, University of Edinburgh, Scotland
Lines: 119

In article <3@tdatirv.UUCP> sarima@tdatirv.UUCP (Stanley Friesen) writes:
>In article <6638@skye.ed.ac.uk> jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
>|I think you are confused.  Here's an example.
>|
>|Suppose we have an argument like this:
>|
>|   1. if P(x) then not can_do(x,task)
>|   2. P(computers)
>|   3. therefore not can_do(computers,task)
>|
>|To show the argument does not apply to humans, it suffices to show
>|that P(humans) is flase.
>
>And that #1 is true. 

No.  To show the argument does not apply to humans it suffices to
show that P(humans) is flase.  Why?  Because then we cannot apply
the rule

   if p then q; p; therefore q

to conclude (3).

In short, here is an argument that concludes computers cannot
understand, and to show it does not apply to humans we need
only show P(humans) is false.  We do not need to show how
can_do(humans,task) manages to be true.

> *That* is usually the undemonstrated assumption in the
>arguments such as the CR.

If (1) is used, there has to be an argument for it, of course.
Nonetheless, to show the whole argument (including the part that
concludes with (1)) does not apply to humans all we have to do
is show P(humans) is false.

>Just claiming that "obviously #1 is true" is not enough, you must provide
>*evidence* that it is true - evidence that is based on *observations*,
>that are *repeatable* by any competent observer.

There are many possibilities between claiming something is obviously
true and providing observational evidence.  Almost all philosophical
arguments fall into the omitted region.

>This has never been done.  In most cases the premise corresponding to #1 is
>true for some set of definitions and false for others, and is thus not
>intrinsically either true or false.  It is *arbitrary*.

What?  The words will be used with some particular meaning, not
as place holders for arbitrary meanings.  Eg, if the task is
understanding, it will be understanding in some sense, not
a completely arbitrary concept.

>Also, it has rarely been shown that P(humans) is false, only that some
>interpretation of human behavior *suggests* it *may* be false.
>
>To show P(human) to be false you must show that the interpretation of
>human behavior that entails it is necessarily true, preferably by observation.

No.  There are many ways to show something is false.  Moreover,
to show something is false it is seldom necessary to show that
something else is _necessarily_ true.  (In any case, if P(humans)
entails Q, and you want to show P(humans) is false, don't you
want to show Q is false (not true)?)

Anyway, you seem to be making a lot of unwarranted assumptions, eg,
that P(humans) entails an interpretation of human behavior.

>In fact, P(human)=false is often entailed by the same definitions that
>make #1 true and thus the two are equivalent (tautology).  Thus little
>is really proven except that it is logically consistant to say that humans
>and computers are different.  This is a *far* cry from saying it is actually
>true of the real world.

You have to remember that (1) is a particular statement.  So it would
be different in a certain respect, not just "different".  Besides, if
it is the case that (1) implies not P(humans), then it is the case
that the argument does not apply to humans.  No one is claiming
that such arguments are therefore correct!  (Only that they don't
apply to humans.)

BTW, they are _equivalent_ only if each implies the other, not if
all we have is that (1) implies not P(humans).

>|To show that P(humans) is false it is not necessary to show _how_
>|humans manage to do the task in question.
>
>No, but it is necessary to show that it is false.

Now you say "no".  So finally you agree with what I've been saying!
Why was it so difficult to get this agreement?

>And in many cases the easiest way to do this is to show that humans perform
>the 'task' in some set of ways.  That is it is easier to go backwards from
>demonstrating the conclusion to the premises than it is to demonstrate the
>premises themselves.

Have I ever argued against doing it in the easiest way?  I don't
think so.

>|Moreover, if we can conclude
>|
>|   can_do(humans,task)
>|
>|we can reason thus:
>|
>|   1. if P(x) then not can_do(x,task)
>|   2. can_do(humans,task)
>|   3. therefore not P(humans)
>|
>
>Again this assumes #1, which is rarely truly *demonstrated* in a
>conclusive way.

Then show it isn't demonstrated!  Do that instead of demanding that
someone else show how humans manage to do the task in question.
That's all I ask.

-- jd


