From: Richard Loosemore (rpwl@lightlink.com)
Date: Thu Aug 24 2006 - 10:21:18 MDT
Michael,
Your comments about the relevance of intelligence are correct: they
simply do not relate to IQ, and the best way to understand this is to
switch to my new, improved test called the "Post Office Quotient".
The POQ is a measure of how far your domicile is from all the City
Central Post Office buildings in your region. It is a U-shaped measure
that goes very low the nearer you get to the building, then peaks at a
distance determined by (among other things) the distance to the nearest
town, and then goes very low again when you get as far as possible from
all the Central Post Offices.
I submit that with this measure you will be able to write a chart that
does not differ substantially from the one you presented:
+1 POQ point = +1.763% income,
Annual gain / POQ point US $55-65 billion
0.4-0.5% GDP
3 point POQ increase:
Poverty rate -25%
Males in jail -25%
High school dropouts -28%
Parentless children -20%
Welfare recipiency -18%
Out-of-wedlock births -15%
If, Michael, you were to concede the point and from this day forth
advocate that we look for people with high POQ in order to staff a
singularity research program, then you would be doing something exactly
equivalent to what you are now doing. I would just be a bit more
obvious that it was nonsensical.
What causes people with high POQ to be so successful, and what causes
people who have the ability to MEASURABLY solve problems more quickly
(or more effectively, for open ended IQ tests) is that POQ and IQ are
correlated with intelligence, which is a cluster of abilities, some of
which we understand and many of which we don't, and some of which we can
measure, and some of which we can not.
In amongst all those abilities that we do not completely understand,
there are some that can be measured. The number you get out of that
measurement process is IQ: it only measures what CAN be measured, and
we have no reason to believe (and plenty of reasons NOT to believe) that
IQ is a direct measure of intelligence-in-the-large. (And please do not
go off on a red herring about the generality of g: those studies, no
matter how comprehensive, are still subject to the criticism that they
measure what CAN be measured, and have no relevance to components that
cannot be measured).
What is distressing (and what illustrates my point about high-IQ people
being just as guilty of closed-mindedness as anyone else) is that you, a
person of supposedly high intelligence, are not addressing the issue I
raised in the above to paragraphs: you seem to be evading it or denying
it for ideological reasons. And just for the record, the summary of the
issue is that
1) There is a difference between correlation and causation, and
2) There is a difference between a partial component of intelligence and
the whole thing, and
3) If a thing can only be partially measured (i.e. only a component of
it can be measured), that fact does not give anyone the licence to
pretend that the measurable bit IS the entire thing.
These things are about empirical science. I truly do not understand why
you (and others) do not seem to either understand them or rank them as
important enough to sway the discussion.
You are not alone in having this attitude, by the way: I have met
psychometricians who have seriously said that because g is such a strong
predictor of behavior, and because there exists such a single factor
that can by a good predictor, therefore g is a measure of THE mechanism
that is intelligence. I have said it to them and I will repeat it here:
those folks are allowing their obsession with statistics and
mathematics to trump science. They are not scientists.
Now, if you were advocating that we look for "intelligent" people to
staff a singularity research program, then, hey, that is what we are all
doing anyhow. It is just that because intelligence is such a difficult
thing to pin down, we know that it would be stupid to take high IQ to be
the only measure of their potential ability.
And as for the idea that IQ per se (rather than intelligence in the
large) is relevant to AGI: sure, it is as relevant as POQ. I suggest
that after we build a candidate AGI we send it out into the world to get
a job, and find a place to live, see where it ends up, and then measure
its POQ. And, yes, that is a little facetious, but you get my point:
concentrating on POQ or IQ now is putting the cart before the horse.
*Intelligence* (and finding out what intelligence actually is, and how
it works) is what we need to be concentrating on.
[Lastly, an aside: I did not accuse you personally of being obsessed by
IQ because yours is (presumably) high, I simply pointed out an
observation I had made of a majority of the people who adulate high IQ].
Richard Loosemore
Michael Anissimov wrote:
> On 8/23/06, Joshua Fox <joshua@joshuafox.com> wrote:
>
>> Might I point out the possibility of one of those good ol' cognitive
>> bias here: Many SL4 members have scored high on IQ and similar tests,
>> and are proud of that, and so may place an unresaonable focus on
>> testing.
>
> This completely untrue. Both Richard Loosemore and now you are
> accusing me of talking about IQ overmuch because of an assumption that
> I scored high and am proud of it. But I know for a fact that there
> are many people on this list with a greater IQ than mine. This silly
> notion - that examining IQ in detail, reading academic papers about
> it, or arguing for its predictive validity are all symptoms of
> irrational exuberance about IQ because of a personal high score.
>
> In fact, I just watched Anders Sandberg's talk at Transvision 06, and
> practically the entire thing focuses on IQ and its importance. He
> cites Linda Gottfredson multiple times, as well as citing the Terman
> study. How come Anders Sandberg can talk for an hour about the
> importance of IQ and everyone is respectful, but when I do the same
> thing I'm being "ridiculous"?
>
> In his talk (available at http://www.cs.helsinki.fi/u/mjrauhal/tv06/),
> he listed many points, including these:
>
> +1 IQ point = +1.763% income (Schwartz),
> Annual gain / IQ point US $55-65 billion
> 0.4-0.5% GDP
> Weiss 1998: 3 point IQ increase:
> Poverty rate -25%
> Males in jail -25%
> High school dropouts -28%
> Parentless children -20%
> Welfare recipiency -18%
> Out-of-wedlock births -15%
>
> Here is a oft-seen informative chart on IQ he presented:
> http://www.acceleratingfuture.com/michael/blog/images/IQ.png
>
> Intelligence is one of the Singularitarian Principles
> (http://yudkowsky.net/sing/principles.html#intelligence):
>
> "At the heart of an appreciation of the Singularity lies an
> appreciation of intelligence. The Singularity places a horizon across
> our understanding because we can't predict what someone smarter than
> us is going to do; if we could, we'd be that smart ourselves.
> Intelligence isn't just the ability to come up with complex solutions
> to complex problems; it's the ability to see the shortcuts, the simple
> and obvious-in-retrospect solutions to complex problems - even
> emotional or philosophical problems. Intelligence isn't just
> high-speed thinking, perfect memories, or other party tricks;
> intelligence is also wisdom, and self-awareness, and other things that
> extend into every aspect of mind and personality."
>
>> Certainly. But how is the measurement of IQ by tests relevant to the
>> development of artificially enhanced intelligence?
>
> Yes. Some of the same features of IQ test problems will be targetted
> by AI designers as tasks that their system must be able to solve.
> Also, a better understanding of human IQ variance in the space of
> minds-in-general will begin with a thorough analysis of that human
> variance. This is important to the argument that human-equivalent AI
> is anthropomorphic. There is no such thing as long-term
> human-equivalent AI: you either have a retard or a superintelligence.
> The transition time from A to B could be measured in the minutes or
> hours.
>
>> However, the discussion in this thread has not been about that
>> angle, but rather about how to best measure the differences between
>> humans in a way that correlates with real intelligence, and this seems
>> only secondarily relevant to SL4.
>
> "It's transhuman intelligence that lies at the heart of the
> Singularity, but we respect intelligence on the human scale as well.
> This respect for intelligence is our shield against being blinded by
> ideology, one of the primary safeguards that prevents
> Singularitarianism from turning into just another banal fanaticism.
> Cloudy thinking in the service of the Singularity wouldn't be a
> virtue; it'd be another dreary manifestation of the same old human
> stupidity that we're trying to get away from."
>
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT
