Re: COMP: Breeding a Better Engine

From: Doug Jones (random@qnet.com)
Date: Mon Jun 19 2000 - 10:10:59 MDT


Ken Clements wrote:
>
> Doug Jones wrote:
>
> >
> > How so? If you're putting forth the squirrel case that Robert Stirling
> > was himself a product of evolution, well then *everything* made by
> > humans is a product of evolution. A tautology of zero value.
>
> Yes, I am, and yes, it is tautology. So is all of mathematics, but that does
> not make it of zero value.

The meta level still doesn't make the program in question capable of
making design innovations, and is useless in the context.

> It is common for folks to make the mental error of
> separating the "works of man" from the "works of nature."

True, but we were discussing the "works of a design evolution program",
running without intervention by man. Human works added to the program
would not be evolution within that system, but would be an external deus
ex machina plunking the program down in a new parameter space.

That program can evolve deisel engines all you want, and it won't come
up with a stirling cycle. If it includes a turbocharger, it could
possibly evolve through a Nene-like stage (a big turbocharged deisel
design of the 1950s) into a jet engine... but the program, as written,
will not invent new methods on its own. Of course, if the programmer
throws in a regenerator, displacer piston, and heat exchangers (along
with the analytical models to predict their behavior), the system might
converge on a useful stirling design- but it could not make that leap on
its own.

Currently available software can do evolutionary design optimization,
but cannot make major jumps to new technologies. Useful, yes, but not
universally versatile. Creative design input is still required.

--
Doug Jones
Rocket Plumber, XCOR Aerospace
http://www.xcor-aerospace.com


This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:29:18 MST