http://www.dailygalaxy.com/my_weblog/2010/05/a-robot-hitler-scientists-show-robots-evolving-to-exhibit-good-evil-a-galaxy-classic.html?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+TheDailyGalaxyNewsFromPlanetEarthBeyond+%28The+Daily+Galaxy%3A+News+from+Planet+Earth+%26+Beyond%29
That's interesting, but I don't think the behaviors qualify as good or evil because I don't think they qualify as learned. They're just machines doing what they're told to do.
ReplyDeleteMeh. Maybe we are too.
Really? I don't think so, I think their behaviour evolved. Who told them to seek food and avoid poison? Nobody told them to do that, "natural" selection favoured those with more charge. I mean, okay, those scientists picked out the ones that with more residual charge to 'breed', but that's as close as you get to a real-life situation where animals with the strength and stamina to put on really impressive mating displays tend to breed more.
ReplyDeleteAh, wait... this is a free will thing, isn't it? You're thinking, each robot can't be held responsible for its actions because it has no free will, and therefore can't be good or evil. I'd argue that good and evil are in the eye of the beholder - it sounds like g & e were pretty good words to describe those actions. If an evil person is someone who performs evil acts, then if a robot performs evil acts, it's an evil robot? Like a man once said, it ain't metaphyics when it's on the table in front of you with an apple in its mouth.
I don't think anyone told the human race to grow up like it has, by the way. I know some may disagree with that. And what are good and evil, anyway? In this research, they sound more like extreme ends of a bell-curve of cooperative responses. And that sounds about right to me, too.
Thanks Andrew. Fascinating stuff!
I said "Maybe we are too." I hedged my bets!
ReplyDeleteI guess I'd inquire about the following:
Do the actions of an "evil" robot benefit at least one other robot? If so, then "evil" is subjective.
The "evil" robots that led the other robots to the "poison" couldn't have done just that, or the behavior wouldn't/shouldn't have been propagated. So they were good at other things too or someone cheated.
The "good" robots that "sacrificed" themselves also displayed a behavior that wouldn't/shouldn't propagate, so there couldn't have been many of them or someone cheated.
It seems the best robot would be one that found the "food" quickly (possibly relying on its fellow robots for help) but doesn't bother sharing (wasting energy passing on the location). Is that optimized self interest good or evil?
When they evolve to the point when a "well-fed" robot pulls/pushes a "poisoned" robot to the food area - thereby changing its alliance - I'll be impressed.
It'd certainly be nice to know whether the all of their behaviour was self-taught... or whether some proportion of it was provided
ReplyDeleteThose are good questions. Maybe it's like plants, which grow to a range of heights even if you cull out the big ones - a population just throws up those extremes every so often as part of its mutations. I don't know. I think the article did imply that the evil robot somehow lured the others into the poison and then scooted off to a feeding station on its own... but as to the martyrs, your guess is as good as mine. Now, there was something on the radio this morning which made me think of this... ah, yes, here it is: are humans more altruistic than bacteria?
ReplyDeleteI think, if you browse my tags, there're a few for ants and altruism
ReplyDelete