From: Charles Hixson (charleshixsn@earthlink.net)
Date: Tue Aug 20 2002 - 16:04:04 MDT
Party of Citizens wrote:
>I would like the interested parties to present a rebutall of the thesis
>developed over the last four days on AI-ARMS-RACE@yahoogroups.com that
>machine intelligence surpasses human intelligence NOW if state-of-the-art
>hardware and software are used by the machine and its intelligent
>performance is rated according to the generally accepted definitions of
>psychologists and educators.
>POC
>
>On Tue, 20 Aug 2002, Amara D. Angelica wrote:
>
>
>
>>http://www.kurzweilai.net/news/frame.html?main=news_single.html?id%3D1191
>>
>>'Cute' robots could take over, warns ABC Nightline
>>
>>ABC Nightline, August 19, 2002
>>
>>With robots getting cuter, like MIT's Kismet and Sony's Aibo puppy, and
>>people playing less with real people and more with fake ones, "one day,
>>adorable robots could do us great harm and we are not ready," warned ABC
>>science correspondent Robert Krowlich on Nightline tonight, August 19.
>>
>>"The people designing these little devices are very, very cunning and
>>exploiting psychology for all they're worth, with faces with winning grins
>>and eyes that they can bat at you," said New School For Social Research
>>psychology professor Nicholas Humphrey.
>>
>>"It's increasingly difficult not to treat these machines as people. They
>>will have their own interests "independent of those that made them, and
>>there will be ways in which the robots will basically get us to do their
>>work for them ... by getting people to form relationships with them to love
>>them.
>>
>>"If one day, they wish to do us harm, we will not be able to resist, because
>>if they assume the form of adorable bunnies or puppies or an adorable human
>>infant, we will embrace them."
>>
>>"We will have to be very clever about how smart we allow them to be or they
>>will succeed humans as the next species running the earth," warned Stephen
>>Petranek, Editor in Chief of Discover magazine.
>>
>>"We'll probably never see it coming," added Sim City game developer Will
>>Wright, whose cute Sim City and The Sims computer games have already become
>>highly addictive with some players.
>>
>>However, artificial intelligence has been a tremendous failure, Petranek
>>claimed. "Robots are really stupid. We've been working on artificial
>>intelligence energetically for about 30 years and now you couldn't find ten
>>people in the world who are working on AI any more because nobody has been
>>able to get very close in trying to mimic the human brain."
>>
>>
First thing I'd do is ask for a definition of your terms. Thus far humans are impressively superior in pattern recognition, goal setting, and operating in an environment that is often wet, dry, or without electricity. Are these intelligence? They are inferior at drawing logical conclusions, and quite inferior at doing modeling of turbulence, trajectories, etc. Are these intelligence? Which?
Both humans and computers are woefully inadequate at understanding the other's language.
My suspicion is that the high home computers are already superior to humans at many of the tasks of intelligence, and that most of the human brain is devoted to such things as keeping your balence, and a lot of it is just oversized because there was the huge cranial expansion, and the brain expanded to fill it. But that a lot of the expansion was from circuitry that was so specialized that it couldn't be freely adapted into general purpose intelligence. Many disagree with me, and I have no evidence, except for the extremely shallow intelligence that people seem to use in many situations. But we seem to have a huge amount of brain adapted for pattern recognition. I suspect that we do both logic and arithmetic via pattern recognition (a recidulously inefficient way to do it, but powerful in wierd ways at wierd places). Computers are so designed that logic is trivial, we had to figure out how to do that, so we were able to tell them efficient ways. But we never figured out how to do good pattern recogniti
on. It's just ... obvious. When computers learn to do that efficiently, they will start to outstrip us in many areas that we currently consider uniquely human. The key, though, is the integration of the various modes of thought. Not just pattern recognition or logic or geometric modeling or goal setting, but some combination of all of these. When this is achieved, then I believe that even with merely modern hardware people would be outclassed. Until then, there is a difference of kind in the intelligence, not merely a difference in amount.
P.S.: Many believe that it would require a much more powerful computer than I feel is necessary to accomplish these ends. They may well be right, only time will show. But when this was accomplished the computer would be more intelligent than humans were, even if extremely slow thinking. Because it could be designed to have thoughts with a dynamically choosen probability of correctness, and people don't get to choose that (at least not very well).
-- -- Charles Hixson Gnu software that is free, The best is yet to be.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:17 MST