http://discovermagazine.com/2008/jan/robots-evolve-and-learn-how-to-lie/

Quote :

"Robots can evolve to communicate with each other, to help, and even to deceive each other, according to Dario Floreano of the Laboratory of Intelligent Systems at the Swiss Federal Institute of Technology. Floreano and his colleagues outfitted robots with light sensors, rings of blue light, and wheels and placed them in habitats furnished with glowing “food sources” and patches of “poison” that recharged or drained their batteries. Their neural circuitry was programmed with just 30 “genes,” elements of software code that determined how much they sensed light and how they responded when they did. The robots were initially programmed both to light up randomly and to move randomly when they sensed light. To create the next generation of robots, Floreano recombined the genes of those that proved fittest—those that had managed to get the biggest charge out of the food source. The resulting code (with a little mutation added in the form of a random change) was downloaded into the robots to make what were, in essence, offspring. Then they were released into their artificial habitat. “We set up a situation common in nature—foraging with uncertainty,” Floreano says. “You have to find food, but you don’t know what food is; if you eat poison, you die.” Four different types of colonies of robots were allowed to eat, reproduce, and expire. By the 50th generation, the robots had learned to communicate—lighting up, in three out of four colonies, to alert the others when they’d found food or poison. The fourth colony sometimes evolved “cheater” robots instead, which would light up to tell the others that the poison was food, while they themselves rolled over to the food source and chowed down without emitting so much as a blink. Some robots, though, were veritable heroes. They signaled danger and died to save other robots. “Sometimes,” Floreano says, “you see that in nature—an animal that emits a cry when it sees a predator; it gets eaten, and the others get away—but I never expected to see this in robots.”"

1/23/2008 8:44:02 AM

neat

1/23/2008 8:53:24 AM

soooo... why couldnt he set it up so they share their own code in some way based on fitness? seemed like he stepped in and did the 'evolutionary' part himself....

i mean his results were interesting but ultimately they didn't do it on their own he picked and choose which parts of the code to pass on with each generation...

it's similar to selective breeding, you can't really say cows evolved to give us that much milk on their own, we limited/encouraged breeding among them to achieve those ends...

good first step though.

1/23/2008 9:04:22 AM

It wasn't about the breeding, it was about the decision making and adapting to lie, help, or merely exist.

1/23/2008 9:13:48 AM

..... yeah, but he could have skipped the 'generation' bull and just started at that point, unless he didn't know which bits would make the best model 'organism'

my point is, from the title and the article, that they didn't really 'evolve' they were bred, there's a important difference.

[Edited on January 23, 2008 at 9:18 AM. Reason : s]

1/23/2008 9:18:30 AM

True enough.

1/23/2008 9:19:42 AM

^^ that's the whole point of using generations. to discover the "best" model

http://en.wikipedia.org/wiki/Evolutionary_algorithm

1/23/2008 9:22:27 AM

coolness

1/23/2008 9:23:43 AM

my issue is with the misconceptions with regards to: evolution vs breeding

the mechanics are almost the same, but with breeding you pick and choose what you want to keep and what you want to get rid of at each generational point, with evolution (alone) fitness alone determines what stays and what gets discarded, not some set of guys.

i suppose when tech takes a bio term they tend to bastardize it a little.

but yes, these sort of methods tend to be far quicker at optimizing, anything really.

1/23/2008 9:28:57 AM

ok i see what you're saying. though it doesn't really say whether he bred or let them evolve. it just says "To create the next generation of robots, Floreano recombined the genes of those that proved fittest—those that had managed to get the biggest charge out of the food source." doesn't say if he evaluated fitness programmatically or selectively

1/23/2008 9:41:32 AM

but he selected and eliminated traits based on his parameters of fitness. pretty much as soon as he stuck hes hands in their genes to recombine them (without having a means or letting them do it on their own) it becomes breeding.

breeding fits inside of evolution but is guided by non-natural (environmental) factors.

but evolution sounds better and elicits more of a response in any sort of public statement.

1/23/2008 9:57:17 AM

bhahahahahahahahahaaaaaaaaaahahhahah

Matt posts a bit of info thats seriously fucking cool and the first reply spends half its time trying to downplay how cool this is.

I mean theres a two things going on here: Evidence of natural selection and evidence that brings light to how intelligence can evolve.

1/23/2008 10:27:10 AM

There are programs for autonomous underwater robots that communicate with each other to defend seaports that are currently under testing...

so combine the 2 and yeah... waternet

1/23/2008 10:33:27 AM

personally, i'm more worried about the other skynet has become self-aware scenario
http://blog.wired.com/defense/2008/01/israel-thinking.html

1/23/2008 10:35:15 AM

And then you've got things like this happening:

http://www.itweb.co.za/sections/business/2007/0710161034.asp?S=IT%20in%20Defence&A=DFN&O=FPTOP

Quote :

"The National Defence Force is probing whether a software glitch led to an antiaircraft cannon malfunction that killed nine soldiers and seriously injured 14 others during a shooting exercise on Friday. SA National Defence Force spokesman brigadier general Kwena Mangope says the cause of the malfunction is not yet known and will be determined by a Board of Inquiry. The police are conducting a separate investigation into the incident. advertisement Click here Media reports say the shooting exercise, using live ammunition, took place at the SA Army's Combat Training Centre, at Lohatlha, in the Northern Cape, as part of an annual force preparation endeavour. Mangope told The Star that it “is assumed that there was a mechanical problem, which led to the accident. The gun, which was fully loaded, did not fire as it normally should have," he said. "It appears as though the gun, which is computerised, jammed before there was some sort of explosion, and then it opened fire uncontrollably, killing and injuring the soldiers." Other reports have suggested a computer error might have been to blame. Defence pundit Helmoed-Römer Heitman told the Weekend Argus that if “the cause lay in computer error, the reason for the tragedy might never be found”. Electronics engineer and defence company CEO Richard Young says he can't believe the incident was purely a mechanical fault. He says his company, C2I2, in the mid 1990s, was involved in two air defence artillery upgrade programmes, dubbed Projects Catchy and Dart. Software details During the shooting trials at Armscor's Alkantpan shooting range, “I personally saw a gun go out of control several times,” Young says. “They made a temporary rig consisting of two steel poles on each side of the weapon, with a rope in between to keep the weapon from swinging. The weapon eventually knocked the polls down.” Young says he was also told at the time that the gun's original equipment manufacturer, Oerlikon, had warned that the GDF Mk V twin 35mm cannon system was not designed for fully automatic control. Yet the guns were automated. At the time, SA was still subject to an arms embargo and Oerlikon played no role in the upgrade. “If I was an engineer on the Board of Inquiry, I would ask for all details about the software for the fire control system and gun drives,” Young says. “If it was not a mechanical or operating system error, you must find out which company developed the software and did the upgrade.” Young says in the 1990s the defence force's acquisitions agency, Armscor, allocated project money on a year-by-year basis, meaning programmes were often rushed. “It would not surprise me if major shortcuts were taken in the qualification of the upgrades. A system like that should never fail to the dangerous mode [rather to the safe mode], except if it was a shoddy design or a shoddy modification. “I think there have been multiple failures here; in software and the absence of interlocking safeguards.” He asks if the guns were given arcs of fire and whether these were enforced with electromechanical end stops. “On a firing range you don't want guns to fire through 360 degrees.” Oerlikon's local agent, Intertechnic, did not respond to requests for comment. The SANDF said investigations were still under way. The air defence artillery will, in the next two years, receive new missiles, radar and computer-based fire control equipment worth R3 billion as part of projects Guardian and Protector."

1/23/2008 12:37:40 PM

interesting bit of info though

1/23/2008 1:19:36 PM

this is cool as hell

1/23/2008 7:59:19 PM

OK, I haven't researched this so it's just off the top of my head but,
I think skynet became "self aware" in 1997, which coincidinkly was the same year
the Robinson family blasted off for alfa centuri in the jupiter 2.
Danger! Danger!

1/23/2008 8:06:19 PM