For those of you that do not know what a FPGA is, it's basicly a electronic chip that has the ability to rewire itself internaly. This allows the chip to recode itself and adapat itself for various situations.
Also this "rewiring" can be very fast, which lends itself very well to neuron evolution, and some people/labs have already started to work on this, and already have a few amazing stories to tell, as you can read from a few threads I started quite a while ago, here and here.
Unfortunately, these things cost as much as a top of the line graphics card, and I have no true electronics training, so I may very well end up trashing the thing, or not being able to use it to its fullest potential...
For an FPGA+CNS system, we would need:
FPGA <--> Results <--> Overseer <--> Mutator <--> FPGA Writer
Basicly, the FPGA does something, and sends its results to the "Overseer", which is a generalistic function, that tries to apreciate if the FPGA is running in the direction ofa solution to the problem with was posed with. The mutator will then mutate the FPGA's structure a bit, and the FPGA Writer will send the new mutated version(s) to the FPGA(s).
This can happen thousands of times per second, and because the neurons won't be evolving inside a computer simulation, but out in the real world, as true electronic components, some very incredible things can happen. If you followed the link I gave above, you've already learned about an FPGA neuron experiment where the evolving circuit inside the FGPA required a clock to function more eficiently, but because none was natively provided, the circuit evolved a sub-circuit which timed and counted electromagntic waves from a nearby diesel generator! It listened to its surroundings, and figured out how to create a clock!
I don't know about you but I find that pretty f**king amazing m8!
Anyways, time allowing, I'll try to have my Robotics website up by today.
Thanks for reading!
Links:  - AI Evolving Circuitry (Zip, 6Mb)