[ExI] What would an IQ of 500 or 1000 look like?

David Lubkin lubkin at unreasonable.com
Tue May 26 02:51:39 UTC 2015

Rafal wrote:

>An interesting question arises: a test capable 
>of discriminating the tails of a normal 
>distribution in a population of 10^30 humans 
>would have to have incredibly tight measurement 
>errors to be valid, i.e. to prevent random 
>fluctuations in scores from throwing up an IQ500 
>result due to luck in answering questions. This 
>means you would need to have many more test 
>items than in a regular IQ test, which may be 
>validated only up to an IQ of 180 or 
>thereabouts. So it could mean that many of the 
>10^30 humans would have to spend weeks on this 
>mega-IQ test before the smartest of them all 
>would be announced to the pan-galactic audience.

Your presumption is that IQ needs to be 
determined through answering a test. There are 
several dozen physical measurements that 
correlate with IQ, e.g., nerve conductive 
velocity, myelin sheathing, etc. We cannot 
directly measure intelligence reliably *yet*. But 
it's not clear there's an inherent reason why we 
couldn't. (And there are also simple subtests of 
regular IQ tests that are good first 
approximations of IQ, like reverse digit span: 
Given a sequence of n digits presented to you one 
per second, repeat the sequence in reverse order.)

Perhaps these measures would supplant traditional 
tests altogether. Perhaps they'd need to be 
normed with test subjects who took your 
über-test. (There's already a Mega Test.) Perhaps 
we'd give the über-test to an overclocked upload 
of the human, for while-you-wait service.

Of course, all this is species-specific. It'd be 
interesting to develop a test that works well 
both for human subjects and for non-human 
subjects with different physical limitations than 
we have, for the pan-galactic, all-species, mental olympics.

-- David.

More information about the extropy-chat mailing list