[UA] CRPG
Epoch
msulliva at wso.williams.edu
Thu Apr 19 13:59:07 PDT 2001
On Thu, 19 Apr 2001, Gareth Hanrahan wrote:
> > Unfortunately, that's what we call the "Turing Test." The ability to hold
> > a conversation in a believable manner is, in fact, what computer
> > scientists consider the only real test for creating a human-equivalent
> > intelligence. The test was originally proposed back in the 1940's, and
> > there have been various attempts at it. Nobody's ever come close. My
> > somewhat informed opinion is that we're a remarkably long way from
> > following it.
>
> There was an article I saw on tuesday (and now can't find the link for,
> proving I'm a moron) about an AI system under development. Basically, they
> took a computer, gave it a fairly good learning algorithm, and sat it down
> with a child psychologist who taught it to speak like she'd teach a baby.
> Apparently it's now got the conversational ability of an 18-month old,
> although how much intelligence it has is debateable.
Ever had a conversation with an 18 month old? :P
I'm dubious of the approach of tossing a learning algorithm on something
and going to town, because my opinion is that we know, pardon the
language, fuck-all about how the human brain learns and stores
memory. What we do know is that it bears damn little resemblence to a Van
Neumann architecture.
Heck, we don't even really know from whence arises the human
intelligence. It's popular for AI proponents to claim that it arises from
the neuron-level of the brain, but we have no hard evidence (that I'm
aware of) for this theory.
Until we do know more about how the human brain works, my opinion is that
saying, "Well, we'll just make it mimic a baby and teach it like we would
a human" is doomed to failure.
Of course, there are lots of very intelligent people who disagree with me.
> Interestingly, one of the project leaders is the guy who did the AI for
> Black and White...
Speaking of which, and dragging this back towards something that resembles
the topic, after the lively conversation on this list about the Sims, is
nobody going to investigate the occult possibilities inherent in B&W?
On another note, the Turing test seems like it's begging for a Major
Charge to be associated with it. Maybe to a deceit-oriented school for
managing to convince a large number of people that a machine has passed
it, or for a technology oriented school for passing it. Surely there's
some kind of UA scenario involving sabotaging advanced AI work to prevent
the Wrong Sort Of People from gaining a Major Charge...
Mike
--
iykwim(aityd), tyvm
- ep
_______________________________________________
UA mailing list
UA at lists.uchicago.edu
http://lists.uchicago.edu/mailman/listinfo/ua
More information about the UA
mailing list