Re: Signal Processing (was BLUE GENE)

From: J. Andrew Rogers (
Date: Tue Apr 11 2006 - 21:54:10 MDT

On Apr 11, 2006, at 7:31 PM, Martin Striz wrote:
> On 4/10/06, J. Andrew Rogers <> wrote:
>> One can do a hell of a lot with remarkably simple primitives. It
>> would be strange to think that evolution created excellent building
>> blocks and then never exploited what groups of those building blocks
>> are capable of.
> Why would that be strange? Evolution has no goals. It doesn't
> implement capabilities that aren't adaptive. I ask you to visualize a
> tesseract.

Evolution has no goals, but it also tends to explore its local space
pretty thoroughly. Capabilities with very high ROI that are
trivially discovered by random walks are things I expect to see
sooner rather than later. That is why it is strange. In case you
have not been paying attention, biology tends to exploit every tiny
sliver of efficiency it accidentally implements with ruthless abandon.

If evolution is driven by fitness functions, why would you not expect
it to find and exploit a cheap capability that can generate very
significant improvements in fitness? Evolution is not efficient or
fast, but when it comes to cheap feature evaluation, it does seem to
be thorough. How many times did it reinvent the eye?

J. Andrew Rogers

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT