[ExI] Off Topic: Turing Test -- ai class at stanford
stefano.vaj at gmail.com
Wed Aug 31 09:52:07 UTC 2011
On 31 August 2011 02:32, Mike Dougherty <msd001 at gmail.com> wrote:
> Right. Some of us could make the case that others here score very
> poorly on the social communication with [normal] humans. The
> disturbing fact about a double-blind Turing test is that people are
> sometimes identified as computers. What does the contrapositive of
> the Turing test prove?
In fact, IMHO, as discussed, the real test is when a number of interviewers
deal with a number of computers of brand X *and* human beings, and the
number of guesses does not exceed what is statistically warranted.
> Also interesting to note that a computer is considered a highly
> successful machine as its number of flops increases, while humans are
> considered much less favorably as their number of flops increase. :)
Very old argument, but "intelligence" has always been identified, both as a
comparative feature of human beings and with regard to AI, as the things
that could not be easily automated at that time (say, ability to memorise
epic poems before their transcription, large integers arithmetics, playing
chess, etc., pattern recognition, etc.).
So, I suspect that IQ tests themselves will have to keep pace with
developments in the AGI camp.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat