3 ms·
I really hope some commentary develops from HN on this
by hammock 11y ago
I really hope some commentary develops from HN on this
- ChuckMcM 11y agoHere is a popular science link to John Koza's version (http://www.popsci.com/scitech/article/2006-04/john-koza-has-built-invention-machine http://www.popsci.com/scitech/article/2006-04/john-koza-has-...) which came out a year before this article. Genetic algorithms produce really interesting results, but like Thompson they also tend to produce irreproducible results, or specifically results that work in the environment where they were evolved but no where else. That makes the output often less than useful and very challenging to certify. Some of the most successful work I know of has been done with antenna design.
- hammock 11y agoThanks. What if you were to try to evolve a program that solves a problem- but tests each evolution simultaneously on multiple chips? Instead of one chip, it must work on 100 chips in order to pass. To sort of cancel out the idiosyncrasies of any one chip. I imagine you'd still have problems but might be mitigated a bit.
- natosaichek 11y agoAdrian ended up modifying his setup later so that rather than running on real hardware, the testbed ran in a hardware simulator that only used the documented properties of the chips. This made the final results not quite as concise (I recall something like 20% more gates used), but they worked on all equivalent chips. I looked for more stuff by him, but he ended up going into evolving database schemas or something like that and got out of the hardware game altogether. Kinda disappointing, really.
- ChuckMcM 11y agoCertainly a two stage fitness process would work, or as mentioned using additional controls on the simulation to avoid idiosyncrasies. It isn't that hard to build a small genetic 'evolver' (my favorites are some of the 'walking' ones which race blob things). But you can evolve all sorts of complex things if you are willing to wait long enough and don't care if you understand the result or not. I built one to evolve a response packet to ssh clients to kill them (I got tired of botnets trying to brute force passwords on a server I ran for a while), all you need is a fitness function, a mutate function, and time :-).
- teraflop 11y agoIt's been posted a number of times over the years but only ever received one comment before this post: https://news.ycombinator.com/item?id=8890167 https://news.ycombinator.com/item?id=8890167