Results 1 to 6 of 6
Thread: Artificial Evolution
July 3rd, 2007 #1
... or one step closer to AI.
Original Link via Pharyngula
Adrian Thompson developed an evolution algorithm that works as an invention machine. The latest in the inventions is a tonal discriminator thingy. The coolest part is this...
Dr. Thompson peered inside his perfect offspring to gain insight into its methods, but what he found inside was baffling. The plucky chip was utilizing only thirty-seven of its one hundred logic gates, and most of them were arranged in a curious collection of feedback loops. Five individual logic cells were functionally disconnected from the rest — with no pathways that would allow them to influence the output — yet when the researcher disabled any one of them the chip lost its ability to discriminate the tones. Furthermore, the final program did not work reliably when it was loaded onto other FPGAs of the same type.
Last edited by LaPalida; July 6th, 2007 at 01:44 AM.Eat, drink, and be merry, for tomorrow we die.
Hide this ad by registering as a memberJuly 4th, 2007 #2
Hmmm good read, I don't know if I would call that AI though. If something is preprogrammed then it isn't really 'thinking' on its own, although thats debatable cause you can say humans are 'preprogrammed' in our own way.
July 6th, 2007 #3
I wasn't thinking of comparing this to humans actually. More to a prion or a virus.
Eat, drink, and be merry, for tomorrow we die.
July 6th, 2007 #4C'est la vie
- Join Date
- Jan 2007
- Thanked 433 Times in 156 Posts
July 6th, 2007 #5Registered User
- Join Date
- Sep 2003
- Thanked 96 Times in 36 Posts
I remember reading about those FPGA's (specifically the one you quoted) 5-10 years ago. I remember one of the things they found with one of the results was that a particular result was actually using the fact that a magnetic field generated by it's own chips altered the way it evaluated the data. Just really crazy shit.
I'm really curious as to the state of FPGA's and evolutionary code after 5+ years of first experimentation and whether anything truly viable has come out of them.
"Every generation sees the past though the lens of its own time." - Thom Hartmann
July 8th, 2007 #6
Haha, that's interesting how the solution adapted to the "quirks" of the individual chip, and also used a very obscure solution with feedback loops and seemingly disconnected parts. I suspect our 'Junk DNA' is similar. When humans build stuff it's very clean, polished and obvious. We favor symmetry and order, probably because it's easier to understand and memorize such constructions.
Since the simulation included "quirks" of the chip, it might not even be deterministic. It could make it harder to analyze things.
Genetic algoritms are always interesting in how they surprise you, and you never quite understand why the heck a particular solution emerged. If you figure it out you smack your forehead and go "Why didn't I think of that!".
A potential drawback to genetic algoritms are local optimums though. Humans have a fantastic advantage: foresight. We can temporarily worsen a solution in order to reach a higher local optimum later on. To combine these two forces of creativity would probably result in... great things.
Jamen jag tror att han skäms, och har gömt sig. Vårt universum det är en av dom otaliga spermasatser som Herren i sin självhärliga ensamhet har runkat fram för å besudla intet.