Slashdot Mirror
Human vs Computer Intelligence
DrLudicrous writes "The NYTimes is running an article regarding tests devised to differentiate from human and computer intelligence. One example are captchas, which can consists of a picture of words, angled and superimposed. A human will be able to read past the superposition, while a computer will not, and thus fails the test. It also goes a bit into some of Turing's predictions of what computers would be like by the year 2000."
I did the gimpy test.
ResultsIt switched pictures on me! Honest!!
We are never going to have a machine that is truly "human". Let me explain.
That doesn't mean we won't have intelligent machines that can do just about anything intellectually that a human can do. A human being is more than just a smart computer. Our behavior is governed not only by the higher logic of our brain, but also by millions of years of bizarre -- often obsolete -- instincts. If you yanked a brain out of a body and hooked it to a computer, it would no longer be truly human because of the lack of hormonal responses that come from every part of the body.
It's simply going to be too hard/impractical and, frankly, useless to make an intelligent machine that mimicked every hormonal reaction and instinctual mechanism.
We will have intelligent machines, but we will never have human machines.
Sometimes it's best to just let stupid people be stupid.
Another area not discussed in the article is vicarious experience, that is, experience and knowledge you have because some cause and effect relationship existed with someone or something else.
For example, the computer's tactile interface has to touch the oven and say 110 deg C, as opposed to taking as fact "I heard a human mention that Unit 5 already did that and it was 110 deg C, so I accept it as fact that it is 110 deg C".
I know I'll get modded down for this, but I wonder what the limits of questioning the computer / human participants was? (Article said they quized participants to see if they could tell who was human and who was a machine). Like, could they ask "What number am I thinking of?" The machine would blank out and the human would stupidly blurt out "69 dude!"
I'd hate to think that computer power didn't increase between the time of Turing's death in 1954, and 2000....
:)
Pretty much any prediction that Turing could make about computers nearly 50 years after his death - and before the advent of transistors - would pure speculation. The fact that Turing's prediction that AIs would be indistinguishable from people in the "Turing test" was wrong, and that other projects based on sheer informational density (such as CYC) have been dismal failures, indicate that it is the purely scripted/explicit logical constraint strategy of solving this problem that is faulty. Unfortunately, the 30 years after that prediction have focused pretty much entirely on scripting and logical constraints, and other methods of artificial/computational intelligence didn't see the light of day until the 80s and 90s.
Be sure to watch further developments in modeling of neurological processes, as there is still hope along this avenue of research
Apparently, there would have to be alt tags that read "Type the word FOO to signify you are a human, not a register bot."
I suppose it could generate a spoken list of words in a sound file that is linked to from the image. The alt tag could then read "Please click to listen to a series of words. Enter the words to signify you are a human, not a register bot."
bytesmythe
Hypocrisy is the resin that holds the plywood of society together.
-- Scott Meyer
CAPTCHAs have several applications for practical security, including (but not limited to): Cool, eh?
What I mean is, I don't think an intelligent being would be capable of creating something that is more intelligent than himself.
The machines need to be programmed by humans, who are limited by their own inteligence.
Can God make a rock so big that he can't carry it himself?
"The majority is always sane, Louis." -- Nessus
http://slashdot.jp
Why don't you devise a test which asks for the sum of 10000 numbers?
Once you devise a test system, someone can write non-AI software that can fake it and pretend to be human by knowing what it needs to for the test. Only a real human can tell human and machine intelligence apart, not a systematic test. That's why Bladerunners had to manually test the androids, instead of just letting a machine do it. Real-time human insight is key to testing machine intelligence.
11*43+456^2
I suppose it could generate a spoken list of words in a sound file that is linked to from the image.
The CAPTCHA web site has such a test, but of the sites that use image-based bot tests, only PayPal offers an audio alternative.
Another problem is that sites often present the tests in proprietary formats with expensive implementation royalties, such as .gif and .mp3.
But even providing both the image in a free image format (.png) and the audio in a free audio format (.ogg) won't help blind users behind a Braille terminal without a speaker, such as blind-deaf users.
Will I retire or break 10K?
Did anyone notice that a lot of these "human" test are also the same ones used for things like hearing/eye tests, color-vision impairment, etc.
This knocks out computers, which lack the intelligence/programming (so far) to differentiate between conflicting objects to make out a letter/numbers.
It also may knock out humans with vision problems though, especially those with colour-vision issues.For those with hearing problems, the sound test isn't good either.
It seems that right now, computers trying to translate these puzzles probably perform along par with old-folks. This also might mean that quite a few seniors may have issues getting a yahoo account though.
I don't believe that Turing proposed the Turing Test as the test to use, but rather as a "mathematical proof" that you could construct such a test.
Basically he said if you could not tell the difference between a computer and a person then you would have to say it was intellignet. ie. this is a way of establishing an upper-bound test - not necessarily that this is the best test.
Unfortunately, IMHO, the AI community and other latched onto this test and put effort into fulfilling the Tring Test rather than more practical and useful goals.
If you asked "Did you sleep well last night?" and the computer said "Me not sleep, me computer." (or some question on some other biological function) then you could probably determine the difference between a human and a computer. This need not, however, preclude machine intelligence.
Engineering is the art of compromise.
Trying the test that was on the NY times article at the original test at
:D
http://www.captcha.net/cgi-bin/pix
I saw turtles. Turtles of which some were swimming. So i typed turtles.
And i FAILED.
"Result of the Test: FAIL
You entered the following word:
turtles
The possible words were:
seashell shell shells seashells"
So, i notice this test does not take into the consideration the limits of second (or generally, non-native) language. English is not my first language and i had seen nowhere that turtles and shells are different?? i saw turtles and some turtles that were in the sea. Turtles.
Uh yea. I take proudly failing in this computer or human test!!! wohoo!!
You can say "there are physical differences between person A and computer B", but the problem is that person A and person C also differ quite a bit.
Saying "foo cannot be done" frequently results in someone being utterly wrong. Just a few decades ago, the idea of atomic power would have been laughable -- the ability to wipe an *entire city* away? How about having a person walk around on the moon? Unthinkable.
So, at the moment it seems to be an insurmountably difficult problem. But, a few years ago, the same thing would have been said about problems that we're not starting to think about being doable via quantum computers -- the face of computer science literally changed.
May we never see th
*note, I'm not sure now that this has anything at all to do with the topic, but it's something that bounces off my head sporadically*
/ex
Recently, I've been working with developmentally disabled people as a job coach-- Making sure they have the ability to do the job they're supposed to, and help them to understand anything that needs to happen.
Part of this is working at a local fast food restaurant. The girl I'm working with can do math fairly well, but she has problems with logic, and pattern matching.
And a few times I started thinking about her as a computer-- She can do math fine, and if I specifically tell her how to match a pattern, she can do it for a short time, but she can't do it in situations like when people order a combo. Let's say they order a #1 with onion rings and a small drink, a #6 large combo, and a kids meal, she won't be able to recognize them as "combos" (she'll read the whole thing back to them item by item.) This brings me to a whole other tangent about user interface design (why the normal methods suck, mostly), but that'll be saved until a proper time.
This has been a difficulty with her position as a cashier, but I find it interesting that I'm more or less programming her as I talk to her and re-affirm her patterns to match.
I wonder if certain disabled humans would fail any "turing" test that were given to them, because they don't have normal pattern matching ability. Furthermore, isn't it possible that instead of trying for fully developed Artificial Intelligence, we should look at perhaps emulating those with disabilities? After all, wouldn't this creation process be easier than a "fully aware", fully pattern realizing person?
AI has always interested me, but I don't know nearly enough about it. The thing that made me notice this is I keep talking to her like I would program a computer ("If this, then that, otherwise this other thing" and "While there is someone in line, take their order").
Maybe I'm off based, or this is already an accepted practice. Can anyone correct me?
Maybe that's the point...a human can (potentially) adapt to any situation or problem. A computer is limited by the confines of its programming.
"You cannot find out which view is the right one by science in the ordinary sense." - C.S. Lewis on Intelligent Design
> According to the UN, 97% of Americans can read.
That depends on how you define "read". Maybe 97% or more of Americans
can read at a basic level, but quite a few of them get lost if you
start using words that are moderately unusual, words with more than
about two syllables, or sentences with more than two clauses, or if
you require a reading speed that approaches the speed at which people
normally talk. I could easily believe 23% can't read in a natural and
easy fashion or read more advanced stuff. I'd be guessing at the
figure, but that sounds pretty close to me. It's worse in some areas
than others, of course. Galion is probably about 20%. The inner
cities tend to be worse.
Also, the percentage who can write coherently is way lower than the
percentage who can read; I would hesitate to call anywhere near 97%
of the population literate if the ability to construct a sentence
and put it to paper is part of the expectation.
Of course, computers write even worse than they read. (If they're
making it up as they go, that is. If they have prefab stuff they
can do pretty well, but that's different.)
Cut that out, or I will ship you to Norilsk in a box.
If that's true, then it really doesn't make sense to attribute Nietzche's conclusions to Nietzche, since they were also deterministic. The Universe figured it out, and Nietzche just manifested a phenomenon.
The counter-argument is that formal systems (such as modern computers) have logical limitations that are not evident in human cognition. Therefore, machines must either make the same leap in complexity such that their actual thought processes can no longer be mapped directly to the underlying formal system, or else remain forever inferior to natural intelligences.
It's also interesting to wonder if Nietzche knew about (or even could have known about) the discovery that nothing is deterministic at the subatomic level. Would he have persisted in his belief that intelligence was deterministic, or would he have theorized that it was probabilistic?
Any sufficiently well-organized community is indistinguishable from Government.
That doesn't really indicate illiteracy. Americans are not used to royal titles. Almost all titles commonly used in this country are usually followed by the surname of the person they describe, so it becomes a habit of sorts to assume that, say, "Dr. Smith's" first name is NOT Smith. Granted, "Charles" is not a name most people would think of as a surname when seen by itself, but I can see how the question might stump some folks if they were asked on the spot and didn't really think about it. (I haven't seen the mentioned show, but I'm guessing it was some guy wandering around the street sticking a microphone in random people's faces and asking them questions...am I way off here? ;) ) Doesn't mean they're stupid or illiterate, it just means that habit dictated their instinctive response ("Um...I don't know...") instead of logical reasoning, which is not unusual in that kind of situation.
DennyK
the question of whether computers use intelligence the same way as humans use intelligence has long been determined through the 'chineese room'.
the point of John Searle's Chinese Room being is to see if 'understanding' is involved in the process of computation. if you can 'process' the symbols of the cards without understanding them (since you're using a wordbook and a programme to do it) - by putting yourself in the place of the computer, you yourself can ask yourself if you required understanding to do it:
Minds Brains and Programmes (The Original Chineese Room):
http://www.bbsonline.org/documents/a/00/00/04/8
the complementary question - 'is the human brain
a digital computer' is answered by the same author:
Is the Human Brain a Digital Computer (John Searle):
http://www.ecs.soton.ac.uk/~harnad/Papers/Py104
Summary of the Argument:
1. On the standard textbook definition, computation is defined syntactically in terms of symbol manipulation.
2. But syntax and symbols are not defined in terms of physics. Though symbol tokens are always physical tokens, "symbol" and "same symbol" are not defined in terms of physical features. Syntax, in short, is not intrinsic to physics.
3. This has the consequence that computation is not discovered in the physics, it is assigned to it. Certain physical phenomena are assigned or used or programmed or interpreted syntactically. Syntax and symbols are observer relative.
4. It follows that you could not discover that the brain or anything else was intrinsically a digital computer, although you could assign a computational interpretation to it as you could to anything else. The point is not that the claim "The brain is a digital computer" is false. Rather it does not get up to the level of falsehood. It does not have a clear sense. You will have misunderstood my account if you think that I am arguing that it is simply false that the brain is a digital computer. The question "Is the brain a digital computer?" is as ill defined as the questions "Is it an abacus?", "Is it a book?", or "Is it a set of symbols?", "Is it a set of mathematical formulae?"
5. Some physical systems facilitate the computational use much better than others. That is why we build, program, and use them. In such cases we are the homunculus in the system interpreting the physics in both syntactical and semantic terms.
6. But the causal explanations we then give do not cite causal properties different from the physics of the implementation and the intentionality of the homunculus.
7. The standard, though tacit, way out of this is to commit the homunculus fallacy. The humunculus fallacy is endemic to computational models of cognition and cannot be removed by the standard recursive decomposition arguments. They are addressed to a different question.
8. We cannot avoid the foregoing results by supposing that the brain is doing "information processing". The brain, as far as its intrinsic operations are concerned, does no information processing. It is a specific biological organ and its specific neurobiological processes cause specific forms of intentionality. In the brain, intrinsically, there are neurobiological processes and sometimes they cause consciousness. But that is the end of the story.
--
best regards,
john
Trouble is we may be able to model it but it doesn't mean we understand it.
At least when we are simulating/modelling weather we can start with base points and do comparisons.
Whereas with stuff like consciousness, I suspect even if the model is broken you might not be able to tell till much later.
If we really wanted intelligent entities which we didn't understand (how they work), there are always humans and other creatures.
The GM bunch may even concoct a few more.