From: J. Andrew Rogers (andrew@ceruleansystems.com)
Date: Wed Apr 12 2006 - 00:15:16 MDT
On Apr 11, 2006, at 10:15 PM, Martin Striz wrote:
> You thought it would be
> strange not to exploit the capabilities of neural ensembles. There
> are lots of potential capabilities that just aren't adaptive in the
> living world, like doing matrix algebra on the fly or modeling
> four-dimensional hypercubes. That's what I was pointing out.
Signal processing is very cheap because it was a trivial exercise to
produce biological hardware with the necessary primitives. Matrix
algebra, on the other hand, is *not* cheap to do biologically except
as the complex aggregate of a boatload of cheap primitives. Delay
lines and summing are obviously cheap in terms of biological
systems. Matrix algebra is not, unless you reduce it to complex
assemblies of the aforementioned primitives (complex assemblies also
being relatively cheap). Wetware DSP was cheap and of high utility,
and therefore a common evolutionary result. If matrix algebra has
utility, such a system is capable of competent approximations.
What would you expect to happen in an evolutionary environment where
biological DSP primitives were cheap and of high utility, knowing
that highly sophisticated DSP processes can be constructed from those
cheap primitives? I think you underestimate the adaptivity of
evolution, particularly when given fine raw materials to work with.
J. Andrew Rogers
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT