[ExI] sciam blog article
John Clark
johnkclark at gmail.com
Sat Mar 26 20:28:08 UTC 2016
On Fri, Mar 25, 2016 Rafal Smigrodzki <rafal.smigrodzki at gmail.com> wrote:
>
> I would expect that a compact description of AlphaGo's underlying
> algorithm is many times shorter than the description of that AI's final
> shape after playing millions of Go games.
>
I would expect the same thing,
and I would expect that the difference in
c
complexity between a humble mediocre master learning algorithm and an
a
astounding genius level
master learning algorithm
to be quite small. There are only about
35 million DNA base pairs
that differ between the genome of a chimpanzee and the genome of a human;
2 bits
per base pair and
8 bits per byte so
there is
less than 9
meg
difference between the two species. And all that
9
meg can't be used for just the improved master learning algorithm, I
believe there may be some physical differences between humans and chimps
too.
I remember
that a little before
the turn of the century I bought a fancy mouse for my computer
that had
a lot of new exotic features like a scroll wheel, but before my computer
could understand the signals from my high tech mouse I had to
i
install some new software, and at the time it seemed
like a
huge
program, it was about one meg. The human master learning algorithm might
not be much larger.
John K Clark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20160326/1252570a/attachment.html>
More information about the extropy-chat
mailing list