I don't go looking for trouble, so I had not realized until now that people are confidently saying that such-and-such an "AI" has some particular level of IQ, apparently as measured by the program's performance in some or another Intelligence Test...
From this is inferred that such-and-such an "AI" has a measured intelligence in (say) the top 10%, or 5%, or 0.01% of the UK population, or whatever...
From which it is stated as a solid objective fact that "AI" is already at "genius level" IQ and will soon surpass all human capability...
After which comes various pseudo-concerned hand-wringings of the plight of human beings in this world where "AI" has inetellectual mastery; or else a techno-triumphalism asserting that humans must now "make way for the rule of machines" etc.
But this is utter nonsense.
That it is nonsense really ought to be obvious*.
If people can't realize this, then I'm afraid they are putty in the hands of those who strategize to destroy us - worse, they are complicit in their own proximate degradation and ultimate damnation.
I feel embarrassed at having to explain (and it is seldom possible to explain the really obvious); but here goes...
"AI" has no "IQ", because it has no intelligence.
None At All.
Computers don't have intelligence.
It is reasonable, and has been done, to measure using IQ tests (specially devised) the comparative intelligence of chimpanzees, or dogs, or other animals - because animals are living beings, and beings have intelligence...
But not computers, because computers don't have intelligence.
Intelligence (i.e. "general intelligence" or "g") is an inferred underlying attribute behind all cognitive abilities - and IQ tests infer this from one or a range of cognitive tests used to put people into a rank order - to which the IQ number is applied on the basis of an assumed normal distribution.
The test is not the intelligence, it is just a way of ranking the intelligence among beings with intelligence.
The IQ test score does not mean anything at all, of itself....
After all, if a person had been provided with a cheat sheet containing all the answers to an IQ test, then the test would would not be measuring intelligence, would it?
Correct answers are not the ability, are they?
An "AI" doing an IQ test is just an elaborate cheat sheet.
Computerized cheating to provide the right answers to a cognitive test is not intelligence.
The validity of an IQ test as a measure of intelligence depends on the fact that the person (or other being) doing the test is actually working it out, by using their intelligence-correlated cognitive abilities.
Computer mechanisms for answering IQ test papers have nothing to do with intelligence - computers don't think; they are nothing-but cheat-sheets when it comes to intelligence testing.
The only intelligence involved in "AI"-cheats on IQ testing; is that of humans; who are needed to provide the computer with the right answers - and programmers who engineer the algorithms for matching the right answers to the test questions.
+++
*I mean that when people are claiming that current "AI" has very high, or above possible, "intelligence" this is a stark reductio ad absurdum - at least for someone capable of understanding the world spontaneously from-themselves and directly-known (rather than through distorting lenses supplied by the totalitarian Establishment).
The answer key has all the answers, but it is not intelligent. AI is cleverly designed that it can regurgitate answers from its training database, but its more like a book than a thinking being. The convenience of having so much information at your fingertips is outweighed by losing the source of the information, the training as to what is relevent and correct being unknown to you, and it tends to generalize, obscuring unique and thus potentially more useful ideas and information.
ReplyDeleteAs such I think AI will turn out to have a lot of use for primary education, but only a very niche use in research and advancing the general frontiers of knowledge and art.
As an office worker it is nice to do the part of the job that is generate BS - such as employee self-evaluations, and to fill out documents. The danger is thinking its better than it is and not keeping a careful eye on it lest the level of garbage gets too noticeable.
@HS: https://charltonteaching.blogspot.com/2025/11/with-ai-whole-paradigm-is-false-and.html
ReplyDelete"An "AI" doing an IQ test is just an elaborate cheat sheet"
ReplyDeleteYes, exactly.
I am not sure exactly how they do it. But it could be something like this: feed solution manuals to IQ tests into the machine, do statistical modelling on those and other texts, and then give the machine novel questions until it answers them correctly. And when it doesn't answer correctly, feed the correct answers into the machine. Of course human beings do not get access to solution manuals nor do they get to keep taking tests again and again until they get the score they want. Which of course is completely fair, especially if the psychologists get paid to turn over the ostensibly restricted solution manuals.
Also, they always sweep the human involvement under the rug. For instance, the famous Deep Blue vs. Kasparov match in 1997 did not show that artificial thinking beats human thinking in chess. What it actually showed was that human beings developed a methodology to select moves in a chess game as well as hardware to implement the methodology. The only thinking was done by human beings.
Likewise, in this case, what has happened is that human beings have developed an elaborate Rube Goldberg machine-type method for imitating correct human responses to IQ test questions.
@NLR - Looking back over the decades, it certainly *looks* like there has been a very long-term and strategic (sometimes unwitting) prepping of Western populations to accept exactly such a combination of "Rube Goldberg machine-type" computation, obscuring the fact of direction by some few humans.
ReplyDeleteAs I've said before, "AI" shares some of the responsibility-hiding, guidance obscuring, functional aspects of bureaucracy and by-vote decision-making.
But neither bureaucracy nor "AI" could dominate as they now do; until our civilization had chosen to embrace sufficient corruption, in several ways; including (inter alia) rejecting God/s, denying our innate wisdom, and abandoning the desire for functionality.