🚀 go-pugleaf

In article <1992Nov4.133442.22560@oracorp.com> daryl@oracorp.com (Daryl McCullough) writes:
>jeff@aiai.ed.ac.uk (Jeff Dalton) writes:
>
>>It's true that I don't know what you look like, haven't mapped your
>>DNA or done other things of that sort; and it's true that the only
>>direct evidence I have is behavioral...
>
>That's all I mean by "using the Turing Test"; using only behavioral
>evidence.

Well, what I mean by it is using a particular form of reasoning.
This reasoning is that _anything_ with certain behavior has certain
properties involving consciousness, thought, etc.  So, all you need
to look at is behavior no matter what else is available.  The fact
that we are willing to make a pragmatic decision when the only
_direct_ evidence is behavioral does not mean that we should regard
such evidence as sufficient in general.

The point about direct evidence is this: we don't forget everything 
else we know (about, for instance, what kinds of entities are capable
of certain behavior) when looking at behavioral evidence.

Of course, so far as the validity of the TT is concerned, it doesn't
matter all that much what people actually do.  What they actually do 
might just be wrong.  

>Whoops. I meant:
>
>     Behavior is the causal relationship between a history of
>     effects of the environment on the system, and the actions taken
>     by the system.

I wouldn't call this the behavior.  Behavior is what something
does, not what causes it, nor the causal relationship between
causes and actions.  Of course, when you observe behavior, you
observe it in some context and in response to various things
that happen.  That's all that happens in the TT.

>>What?  Are you suggesting an intelligent entity cannot behave in
>>nonintelligent ways?
>
>Yes.

An intelligent entity can behave in unintelligent ways in a TT,
why not?  It can just pretend to be stupid.

>    As I said, behavior is a causal relationship between inputs and
>outputs, it is *not* simply a sequence of actions. While it may be
>possible to classify actions as "unintelligent", and it may be
>possible for an intelligent entity to produce an unintelligent action,
>I don't believe it is possible for an intelligent entity to have an
>unintelligent behavior (or vice-versa).

I think it's virtually meaningless to talk about an intelligent
or unintelligent causal relationship.  But if you want to go that
way ...

What you seem to be up to is another one of these theory-saving steps
(like "disposition").  The causal relationship can be "intelligent"
even if this "intelligence" is never seen in any actions.  However, 
if this is what you mean by behavior, the TT no longer tells you
wnat the behavior is, and so it's relevance is at best obscure.

>>And if not, what has been ruled out?
>
>You can rule out the possibility that the entity completely ignores
>what is said to it (as a phonograph does). 
>
>You can rule out the possibility that the entity simply repeats what
>is said to it (as a parrot does). 

At least we have falsification now, instead of verification.

>You can rule out the possibility that the output
>is simply a syntactic rearrangement of the inputs (as Eliza does).

How do you rule out that it's doing a more complex syntactic
processing?

>>>It's an imperfect judgement based on insufficient data, but then so
>>>are all judgements that people make in the real world.
>>
>>But we don't treat them all as equally justified!
>
>Agreed. But the judging intelligence (and other mental properties)
>based on behavioral evidence is about as justified as any fallible
>test can be.

I disagree.  Moreover, I have yet to see anything even approaching
a reasonable justification for it. 

-- jd