gcj%qmc-ori.uucp@CS.UCL.AC.UK (03/20/86)
``How a pair of dull-witted programs
can look like geniuses on I.Q. tests.''
This article appeared in the March issue of Scientific American
in the Computer Recreations column of A.K.Dewdney which discusses
the concept of an IQ test for computers, (cf Vol 3 # 164 et seq).
He mentions the HI Q program of Marcel Feenstra, which solves
problems of the "sequence completion" and "numerical analogies"
types. This scores 160 on the corresponding parts of the IQ tests
described by Hans J. Eysenck. Dewdney describes his own putative
program SE Q.
Dewdney paraphrases ``The Mismeasure of Man'' by Stephen J. Gould
and says :- ``What it comes to is this: The traditional I.Q. test
rests on the unstated assumption that intelligence, like strength,
is a single quality of human physiology that can be measured by a
graded series of tasks.''
So far, so good.
He then quotes Gould directly :- `` Our brains are enormously
complex computers''.
Hmmm... getting a bit fishy.
Finally, he says :- `` Does the score on the test measure the
intelligence of the computer? If it does not, just how does one
go about measuring the intelligence of a computer, whether it is
made of silicon and plastic or carbon and tissue? The answer:
Probably not by running some I.Q. program through a battery of
tests.''
Two gripes with this. Who are the carbon/tissue *computers* he is
talking about? Secondly, computers will never be "intelligent";
however software might *appear* intelligent in certain respects.
Nuff said.
Gordon Joly
ARPA: gcj%qmc-ori@ucl-cs.arpa
UUCP: ...!ukc!qmc-cs!qmc-ori!gcj
P.S. Funny, I thought the Answer was 42.
`` The monkey spoke!'' - Zaphod Beeblebrox on Arthur Dent.