XPost: rec.arts.sf.written, rec.arts.sf.science
From: robban@clubtelco.com
On 27/01/2014 11:59 pm, Gutless Umbrella Carrying Sissy wrote:
> Robert Bannister wrote in
> news:bkj8gvF3bu8U1@mid.individual.net:
>
>> On 26/01/2014 7:26 am, Gutless Umbrella Carrying Sissy wrote:
>>> JRStern wrote in
>>> news:1238e9dot92ieomstdc66qkq2g0hs0nqkf@4ax.com:
>>>
>>>> On Fri, 24 Jan 2014 15:47:53 -0700, Gutless Umbrella Carrying
>>>> Sissy wrote:
>>>>
>>>>> Your Name wrote in
>>>>> news:250120141214165689%YourName@YourISP.com:
>>>>>
>>>>>> In article , Ryk E. Spoor
>>>>>> wrote:
>>>>>>> On 1/24/14 3:13 PM, Gutless Umbrella Carrying Sissy wrote:
>>>>>>>
>>>>>>>> The point Greg made, that has not been disputed in any
>>>>>>>> way, is that the definition used to be "do only one thing
>>>>>>>> well," until computers could do only one thing really
>>>>>>>> well, and now the definition has changed.
>>>>>>>
>>>>>>> Oh, I would DEFINITELY dispute that. AI was always "A
>>>>>>> machine that
>>>>>>> thinks like a human, only maybe better", and the Turing
>>>>>>> Test (as a general concept -- making one that really works
>>>>>>> is harder) was always the general idea of how to really
>>>>>>> measure it. Can it pass for human in realistic
>>>>>>> circumstances?
>>>>>>>
>>>>>>> The fact that the concept is as foggy as our
>>>>>>> understanding of what
>>>>>>> intelligence IS is what causes the confusion.
>>>>>>>
>>>>>>> The definition of "thinks like" has been refined through
>>>>>>> the years,
>>>>>>> yes. And people -- usually laymen -- would put up examples
>>>>>>> of tasks that "only a true AI could solve!", like chess,
>>>>>>> but anyone with any skin in the game knew that this wasn't
>>>>>>> true; enough brute-force would beat any human without any
>>>>>>> actual intelligence involved. It WAS thought that computers
>>>>>>> would never HAVE such brute force available and that,
>>>>>>> therefore, any computer that could do grandmaster chess
>>>>>>> must be doing something intelligent, but Moore's Law
>>>>>>> changed that.
>>>>>>
>>>>>> "Artificial intelligence" has never been defined as the
>>>>>> ability to do just one particular thing well, no matter how
>>>>>> complex that thing may seem to be.
>>>>>>
>>>>>> Even ignoring Turing, the true test of "artifical
>>>>>> intelligence" would require the ability to do many, many,
>>>>>> many different things, and to be able to learn to do new
>>>>>> things.
>>>>>>
>>>>> Is driving a car "one thing"? Or is it many, many things, all
>>>>> performed at the same time? How about "driving and navigating
>>>>> at the same time, while also mointoring the physical
>>>>> condition of the vehicle, fuel usage, and driving
>>>>> conditions"? Is that one thing, or several?
>>>>>
>>>>> You can't even come up with an meaningful, objective
>>>>> definition of "one thing." And you still haven't addressed
>>>>> Greg's point, that once you do, when computers can do that,
>>>>> the definition will be changed again, so that computers
>>>>> cannot, by (meta) definition _ever_ be "intelligent," because
>>>>> the people doing the defining will simply not stand for it.
>>>>
>>>> What you said.
>>>>
>>>> But let me rant some along these (familiar) lines.
>>>>
>>>> One of the first uses of the term AI was that it was going to
>>>> take "intelligence" for a computer to juggle multiple threads.
>>>> Now that is entirely a commonplace, done with a few lines of
>>>> code.
>>>>
>>>> Chess is still on the bubble here, originally it was said to
>>>> require so much creativity and insight it was unthinkable that
>>>> a machine could ever do it. Of course such statements showed
>>>> gross ignorance of mathematics and game theory, or at least a
>>>> pessimistic view of how fast computers could ever get.
>>>
>>> Indeed. Nobody worth taking seriously ever said it was
>>> impossible.
>>>
>>>> Fifty
>>>> years later chess grandmasters still go on about how they hate
>>>> being beaten by a machine that uses just pure search, but if
>>>> they add a few heuristics it looks so much more "human" and
>>>> "creative" and that makes it better. Hmm.
>>>>
>>>> So maybe there is a tad of truth to the idea that some stuff
>>>> can be labeled as "requiring intelligence", and then it turns
>>>> out, "oh, maybe not so much after all." But what it suggests,
>>>> frighteningly, is that ALL things eventually fall to that
>>>> process.
>>>
>>> There are those who argue that the human brain is, in fact, a
>>> deterinistic computer. The more we learn about intelligence,
>>> artificial and otherwise, the more firmly they believe that.
>>>>
>>>> I think the "truth" is probably a little more tricky than that
>>>> and both sides are a little bit right, say you had a program
>>>> you could put into your PC and it acted just as intelligently
>>>> as anything else you can communicate with via computer. Maybe
>>>> it really is as "intelligent" as you, as feeling, and as
>>>> screwed up. But, um, what then? Good topic for some scifi
>>>> stories, boatloads of which have already been written. From
>>>> scifi, truth.
>>>>
>>> There are also those who believe that there is more to the
>>> human brain than the organic equivalent of transistors, and
>>> that it is not purely deterministic. Some talk about quantum
>>> effects, some talk about "soul," but notobyd has come up with
>>> anything that can be measured so far.
>>>
>>
>> There has been a lot of work to determine how various chemicals
>> can affect how and even what we think. We may discover that free
>> will is an illusion.
>>
> If it is an illusion, then it seems likely that we _must_ disvoer
> it, doesn't it?
>
I think so, though I dread to think of the uses the military could put
it to.
--
Robert Bannister - 1940-71 SE England
1972-now W Australia
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
