[ExI] Fw: Re: atheists declare religions as scams.

Dave Sill sparge at gmail.com
Sat Jan 29 23:21:16 UTC 2011


2011/1/29 spike <spike66 at att.net>
>
>
>
> (1) to learn, (2) to understand, (3) to make inferences.  Two of these have
> been accomplished and one is difficult to define.  A chess program which
> keeps a memory of which openings it tried and how the game ended, then makes
> adjustments to its own play is an example of (1) learning.
>

That's a chess-specific example of learning. I'm talking about general
learning. Can one teach Deep Blue to play checkers? Nope. Then Deep Blue
isn't really intelligent.

  Done.  (3) Making inferences (depending on how it is defined) is how
> Watson operates when playing Jeopardy.
>

Again, that's domain-specific. If I can sit down and chat with Watson,
explain things, ask it questions that require it to make inferences, and get
answers that demonstrate that it has done that, then I'll count it.


> Two down, one to go.  (2) To understand.  Hmmm, to understand.  Well you
> might have me on that one.  I would be tempted to point to humans who
> clearly do not understand.  I don’t understand a lot of things that others
> get: human emotions for instance.
>

I didn't say "understand everything". There's a wide range of understanding
of an enormous range of topics demonstrated by the 6 billions proles on the
planet at the moment. None of them understand everything and all of them are
human-level intelligent.

  To argue that (2) to understand is a necessary requisite for intelligence
> requires further definition.
>

http://en.wikipedia.org/wiki/Understanding


> >… We may not understand the process today, but I don't think it's beyond
> our comprehension…
>
>
>
> Agreed.  I argue that it may be remarkably difficult to recognize
> intelligence if we see it.  We may not understand understanding.
>

I don't think we know everything there is to know about it, but most can
tell pretty quickly one someone else doesn't understand something.


> >…BTW, where's this chat-room Eliza that passes the Turing test? –Dave
>
>
>
> I don’t know.  Does anyone have that?  Last time I went looking for it
> online, it was gone.  It’s been at least 6 years ago.  A professor rigged up
> a specialized version of Eliza and set it to go into a teen chat room.  His
> reasoning is that most people who went to college in the 70s or 80s probably
> played with Eliza, but those who were born after about 1985 might not have
> even heard of it.  Sure enough, most of the teens chatted away with it for a
> while before becoming suspicious.  Many of those who did figure it out did
> so because the responses were so fast.  A lot of the entries were “Damn you
> type fast.”
>

That's a cute story about some people who were duped by a chatbot. That's
not exactly a Turing test, where an interviewer is chatting with two
entities, one human, one AI, for the purpose of determining which is which.
Fooling unsuspecting people is *way* easier than fooling a competent,
informed interviewer.

-Dave
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20110129/b23c839f/attachment.html>


More information about the extropy-chat mailing list