6 ms·
If you want to read what I think is one of the coolest experiments with GAs I've ever seen, check out the paper "An evolved circuit, intrinsic in silicon, entwi
by mk_stjames 3y ago
If you want to read what I think is one of the coolest experiments with GAs I've ever seen, check out the paper "An evolved circuit, intrinsic in silicon, entwined with physics." by Adrian Thompson [0] - taking the bitstream of an FPGA and running it through a GA and them testing the hardware running in-circuit while judging fitness to a desired function. Crazy things ensue.
[0] https://link.springer.com/chapter/10.1007/3-540-63173-9_61 https://link.springer.com/chapter/10.1007/3-540-63173-9_61
but likely can be found online in other places
- Tommstein 3y agoI remember presenting that paper in grad school. Can confirm the ensuing WTF factor.
- flir 3y agoThat paper, and TS Ray's Tierra, left me with a life-long fascination with GA. But as I understand it, as long as you're chasing a fixed fitness function the technique doesn't really outperform simulated annealing.
- markisus 3y agoI just looked up this paper from "other places". The author evolves an FPGA to discriminate between 1Khz and 10Khz square waves. I don't know that much about FPGAs, but it seems like such a task should not be possible without an external clock signal, since otherwise the output wave-form the 10Khz signal should simply be the 1Khz signal sped up 10x. Somehow, evolution learned to make use of electrical and thermal effects to accomplish the task. There were even units that were electrically separated from the rest of the device which were crucial to its functioning. Pretty amazing.