urth-urth.net Digest, Vol 5, Issue 41

Chris · 28 Jan 2005, 14:30 · from 2005-January.txt.gz

This is the message as the crawl holds it, before anything here touched it. What the conversation shows has been through ingest: transfer encodings decoded, attachments dropped, quote markers repaired, and the archiver’s ? resolved into quotation marks and dashes. Those repairs are judgements and some of them are wrong, which is what this page is for. Email addresses are obscured here as they are everywhere else on the site, so it is before our repairs rather than byte for byte.

From: rasputin_ at hotmail.com (Chris)
Date: Fri Jan 28 14:30:07 2005
Subject: (urth) Re: urth-urth.net Digest, Vol 5, Issue 41
In-Reply-To: <1f76173705012813136be35e88 at mail.gmail.com>
Message-ID: <BAY102-F15FBFE80FEC922725154B786790 at phx.gbl>

Dan'l said:
Turin said:
> > the imitation game taken literally is the same kind of reductionism
> > as behaviorism.
>
>The imitation game stands as the only empirical test thus far
>proposed for identifying a machine intelligence, if one should
>arise. It has several flaws: I find most significant among these
>its failure to account for the probability that the experiential life
>of a machine intelligence would differ in important ways from that
>of a human; thus, it would tend to eliminate a (hypothetical as of
>today) class of genuinely sentient mechines which were not good
>at imagining and empathizing with the experiential life of humans.

I think it's possible to accept the test without engaging in full-blown 
reductionism, although yes, it does smack of behaviorism regardless. 
Basically by endorsing the test you are making an ought-claim (possibly of 
the ethical variety, although the ethical isn't the only type of "ought"): 
that, regardless of the truth behind the currently impenetrable curtain, we 
*should* make our judgments concerning intelligence based on evidence of 
behavior. It is, granted, a short step from there to saying that the 
evidence of behavior *is* the only truth of the matter there is, that 
there's nothing to see behind the curtain - but this additional step isn't 
really necessary to accept the test for what it is.

I have often thought about the particular flaw that Dan'l identifies. It 
seems to me that the class of sentient machines that you're saying would be 
excluded would really be comparable, all in all, to madmen. A madman might 
have trouble passing as "sentient" by rigid standards, and yet they are - at 
least to some extent - somehow identifiable as intelligent, but insane. 
Labeling a machine as "insane" might be quite unfair, but pragmatically 
would have to be done. A machine intelligence with perceptions and 
motivations sufficiently different from our admittedly arbitrary norm would 
be far more dangerous than any homicidal madman.

> > the test itself was just the "chinese room",
>
>You betray your prejudices by even mentioning Searle's quibble.

I really don't follow what you mean by this. He may have been mentioning 
"Searle's quibble" because I brought it up earlier, and I don't think that I 
have a bias on the matter, nor do I think simply mentioning him in this 
context is irrelevant or biased.

-- Civet

--
"You must not be soft, but cover yourself up; for you must discover an 
intellection abstractional and fraudulent."

An unofficial mirror of a public mailing list · about & removal requests · download the archive · before the list · original archive at urth.net