[extropy-chat] Extreme Intelligence

[Lee Corbin]

Samantha writes

> > Could an intelligence exist using our basic human architecture
> > that could rapidly solve problems far harder than anyone can
> > solve today?
> >
> > Here is what I have in mind: suppose first that there is a
> > canonical way to extend the IQ scale. Then would it be possible
> > for a set of atoms to exist, human in form, such that using our
> > same sense organs and with a brain less than twice as large as
> > ours it would have an IQ high enough to in one week flat
> > accomplish any of the following?
> 
> What do you mean by "human in form"?  Looks like a human and acts  
> like a human and has human DNA, etc but is much smarter?

Right.

> If so then adding circuitry and deep connectivity to a human brain, preferably  
> within the skull, should allow quite a bit of improvement.  Replacing  
> select parts of the brain with electronic faster equivalents is one  
> way to do it.

I did want to forbid simple speed-up with, say, electronic equivalents.
But your remarks on deep connectivity open an aspect I hadn't considered.

> >   * figure out how a fusion energy reactor could be designed,
> >     and write up specifications sufficiently detailed so that
> >     the rest of us could build the thing
> 
> Writing detailed specs is not a function of only greater  
> intelligence.  It takes slow meat world time.  So no unless
> the uber-human is equipped with direct computer/brain IO.

If you and Russell are right in your negative assessments, that in
itself is quite remarkable: To wit, that no matter how intelligent a
human being could be, it would not be possible for him to complete this
task in the allotted time. (Almost every sentence with a "no matter how"
or an "every" is pretty remarkable, if true.)

Consider the admittedly far more abstract task of outlining a proof
of Fermat's Last Theorem that human mathematicians of 1970 could
have used to construct an actual proof. I dare say that there exists
a two-page description that the 1970 mathematicians could have read
and understood, a description that would have allowed them a straight
path to a proof along the 1994 lines that they could have implemented
in a few months.

Think about all the descriptions limited by (parametrized by) number
of words that exists in the abstract; for example, the set of all 500
word descriptions. Such a concept can be useful. Here is one application:
In philosophy, Chalmers and others consider the "hard problem" of human
consciousness. But I do not think that there can exist any combination
of 500 (or 5000) words that would supply the answer they're looking for,
and that the reason is that they are simply looking at the problem wrong.

But I do think that there exists a 15,000 word outline of how to build
a nuclear power plant. So you are saying that no conceivable human would
likely stumble upon that description.

> >   * be able to understand at about normal reading speed any
> >     book ever written, much as you can easily understand and
> >     absorb everything being conveyed by a Dick-and-Jane book
> 
> Yes and no.  Order of new knowledge would be important due to  
> dependencies.  With that caveat and the caveat that actual
> learning and retention were similarly enhanced, yes.

And of course, the proposed activity of a super-smart human in
devising an outline of Fermat's Last Theorem is an application
of being able to read and understand almost anything. I figure
that it was physically possible for some collection of molecules
(in the shape of and with the mental architecture of a human)
to have in 1970 thought about Fermat's Last Theorem for a few
days and written up such an outline. The emphasis is on *some*
collection, as the emphasis is supposed to be on "Extreme" in
the subject line of these exchanges.

Lee