From: Yan King Yin (y.k.y@lycos.com)
Date: Sun Jan 11 2004 - 20:22:02 MST
From: "Ben Goertzel" <ben@goertzel.org>
>> From: "Metaqualia"
>>
>> My current working hypothesis is that the complexity and richness
>> of qualia
>> corresponds directly to the complexity of the algorithm.
>
> [...]
>
>The big question is: What is the right definition of "complexity" here?
>That will only come out of a theory that crosses the experiential and
>physical perspectives on mind ... i.e. a type of theory that we don't really
>have it ... but presumably will have one day...
I totally agree.. complexity of the brain is an extremely
important topic...
First, the brain's complexity, *excluding memories that came
from experience*, is actually about 100K genes, which sets
the upper bound of its algorithmic complexity. I'm not very
familiar with complexity theory but am under the impression
that all common measures of complexities are about the same
"big O" order. Which means the information entropy of the
brain (minus memories) should also be less than ~100K.
This means that the brain is actually not a very complex
program (by current definitions of complexities). But it's
also a program that took billions of years to write (evolve).
What we're interested in is when will AGI be possible. We can
identify several limiting factors: 1) raw computing power (as in
teraflops per second); 2) memory capacity and 3) bandwidth; and
4) algorithmic complexity. All of these must be matched before AGI
can emerge. We may predict the first 3 can be reached due to
Moore's Law and the like. Algorithmic complexity of the brain
seems to be a trivial factor. But we're aware that AGI may still
be hard even after all 4 factors are met.
What may be missing is the space/time resource required to
write the AI. Does a program that took billions of years to
evolve require proportional amounts of time to immitate? What
is the appropiate theory to address this issue?
YKY
____________________________________________________________
Get advanced SPAM filtering on Webmail or POP Mail ... Get Lycos Mail!
http://login.mail.lycos.com/r/referral?aid=27005
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:45 MDT