From: Michael Vassar (michaelvassar@hotmail.com)
Date: Fri Nov 25 2005 - 00:18:45 MST
The Feynman, Watson, and Crick IQ numbers are obviously bogus. Watson was a
freakin Child prodigy, and Feynman was a Jew at MIT back when they had
quotas about accepting too many Jews, and back when IQ tests were still
primitive. Even if the number isn't a simple lie, you can bet that the MIT
admissions committee did a better job measuring 'g' than the tests of the
day. As for Crick, have you actually read him? Simply the quality of vocab
and grammar would put him well above a 115 on tests which are substantially
tests of vocab. Obviously, he was a physicist, which demands (as in getting
through a single physics course with decent grades is impossible
without)quant and spacial skills which would lead to better performance on
those sections of most tests. People like to make up factoids.
Anyway, Mark Geddes, since you wonder how Eliezer would have done on that
puzzle, I'll tell you exactly how he would have done. He would have said
something like "there is a minute chance that someone working on such a
puzzle is smart enough that they could contribute to a seed AI, but no
chance at all that someone working on such a puzzle rather than on something
useful has the minimal level of ethical seriousness that must be demanded
from a seed AI researcher so it isn't worth recruiting here." As a serious
seed AI researcher, the possibility that it might be a good idea to spend
his time working the puzzle would never even have occurred to him. People
pay attention to Eliezer not because he's an ex-prodigy and supposed
"super-genius", but because he has spent his time effectively to develop
certain ideas with more clarity than anyone else has. Your ability to make
accurate one sentence guesses, without the ability to explain your
reasoning, is just not interesting, even if it does result in your scoring
higher on some IQ tests than Eliezer or anyone else would. Maybe the puzzle
you were talking about was "the ultimate IQ test" but that is precisely why
it is NOT similar to FAI theory. Unlike the answers on an IQ test, "AI
doesn't fit on a t-shirt", and for FAI, which can't be solved with guess
work, trial, and error, even correct intuitive answers which cannot be
explained are little better than worthless.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:53 MDT