Eugene Leitl wrote:
> a biological organism,
> comparing to our trivial handiwork is the ultimate in complexity and
> robustness, and is not the product of rational design. As to
> work (computation), that is cheap enough with MNT.
> Hijacking that principle for technical designs appears perfectly
> possible, in fact this is exactly what's happening right now, albeit
> the results have not spilled into the industry yet..
> Of course the artifexes, unless cheaply cloned, are the bottleneck in
> the equation. The point is, you don't need them. YMMV..
> You could do a lot of funky
> things though, if you'd apply GA to a really fast simulator engine
> (let's say a throughput of about 1 gps (generation/s) on a population
> of a few thousand individua)..
To do evolutionary design you need three things: a mechanism for simulating (or creating) a large population of possible designs, a mutation mechanism capable of producing useful variations on an existing design, and a pruning mechanism capable of weeding out inferior designs.
In the biotech field, where people are using these techniques to make molecules, meeting these conditions is easy. Real molecules can be built and tested quickly by automated equipment, and generating vast numbers of variants on a known molecule is fairly straitforward.
In the engineering world, none of these conditions can be met. You can't build a machine that will simulate ten thousand variations on a next-generation processor at a high time rate, and building real processors instead is too slow and expensive to be useful. Even if you could, making ten thousand random changes in a chip design will simply give you ten thousand non-functional chips - the capacity for incremental improvement that biological organisms exhibit simply isn't there. Finally, how do you tell which design is best? A simple-minded speed test won't do - you need a complex set of performance measurements, and an even more complex set of tests to make sure there are no hidden bugs, all of which would have to be reviewed by humans.
In principle you might be able to make devices capable of evolution, but you would sacrifice performance and predictability to achieve that goal. Even then, it is unlikely that you could evolve useful nanotechnology as fast as it will be designed - evolution simply takes far more processing power than intelligent design.
> Hardware as http://18.104.22.168/mpf/Nano97/paper.html , you mean?
> Narrowing down the design space a lot further does not appear exactly
> impossible. If the omega hardware is not the proper hardware for the
> SI to run on, what else?
The paper you reference is an interesting theoretical study, but it contains nothing resembling a blueprint. "Blue sky" theorizing is not the same thing as detailed design work. Turning the idea into reality requires an engineering project that would make the Apollo program look like a couple of kids playing with toy blocks.
> I could initiate a search in rule
> space and cast the optimal or near-optimal rule into molecular
> circuitry with comparatively little effort (well, make that a
> future me in a few decades). If your computer is a 3d array of
> identical, simple cells, all you have to worry are details like
> power, cooling, and I/O. Unless you're a pioneer with a faible
> for herculean deeds, you can reuse parts of other designs..
If it were that easy, we'd already do it that way. For almost any problem of significance, the applicable rule space is far too large to be searched by any simple-minded process. If you don't believe me, give it a try..
The bottom line is that in engineering, there is no magical substitute for sentient minds. Turning an idea into a functioning device is a process that requires creativity, and the chances of it being automated without AI are pretty slim.
Billy Brown, MCSE+I