Slashdot Mirror
Evolving Robots Learn To Prey On Each Other
quaith writes "Dario Floreano and Laurent Keller report in PLoS ONE how their robots were able to rapidly evolve complex behaviors such as collision-free movement, homing, predator versus prey strategies, cooperation, and even altruism. A hundred generations of selection controlled by a simple neural network were sufficient to allow robots to evolve these behaviors. Their robots initially exhibited completely uncoordinated behavior, but as they evolved, the robots were able to orientate, escape predators, and even cooperate. The authors point out that this confirms a proposal by Alan Turing who suggested in the 1950s that building machines capable of adaptation and learning would be too difficult for a human designer and could instead be done using an evolutionary process. The robots aren't yet ready to compete in Robot Wars, but they're still pretty impressive."
What could possibly go wrong?!
Minor detail perhaps, but as Academic Editor in Chief of PLoS Biology I want to point out that the paper was in PLoS Biology not PLoS One ...
And those who aren't saved go to robot hell, and must play the fiddle and beat the robot devil in order to leave.
Hey now. I for one welcome our new robot...oh fuck it. It's been done too many times.
actually, as of now, these robots are just programs in a physics computer experiment... so if they were to evolve to be smart, we'd have a computer virus instead of an actual robot that is evolving. I wonder, if a robot program like this were let loose on the internet, and was capable of learning... what would it learn?
Well thank god we don't have to worry about that then, we can win the war by trapping the little fucker in IE6.
Definitely an interesting continuation of work being done by various groups over the past couple of decades.
But one thing to note is that crossover isn't especially useful in neural network evolution. In early stages of evolution, it's really no better than random large perturbation of large swaths of the genome. In later stages, it can actually decrease the speed of evolution toward high fitness genomes, because at least some of the time (particularly if there are multiple "species" in the population) crossover ends up being a random large perturbation which hinders the search of local fitness space by mutation; the rest of the time (when individuals from the same "species" are crossed) crossover is no better than mutation.
The reason for this is because the parameters of a neural network are not functional. A section of the genome may correspond to a weight between neurons, but that weight doesn't have a specific function. In biological organisms, each gene is transcribed/translated into a protein, and that protein may have a particular function within the cell. If that gene is acquired by a descendant through crossover, the protein could serve the same (or a somewhat modified) role it served in its parent, even if the rest of the descendant's genome was acquired from the other parent. But with artificial neural networks, the parameters were all evolved as parts of a whole, where each individual parameter has no function on its own, but the behavior emerges from having all of those parameters at the same time.
This could potentially be mitigated by the genome encoding scheme one uses, and of course, if the crossover rate is low enough, the ultimate effect would be small.
The noun "orientation" is derived from the verb "orient", not the other way around.
-- "At Microsoft, quality is job 1.1" -- PC Magazine, Nov. 1994
The predator and prey bots reminds me of sales people chasing around after anyone who wanders too closely while they try their sales pitch.
This kind of behavior was first demonstrated/modeled (AFAIK/IIRC) as part of the Tierra simulations almost twenty years ago. Though I don't have a reference to hand, I know it's been done in neural networks before too.
So other than the 'sizzle' (as opposed to 'steak') of doing it with robots, can anyone explain what is new here?
You don't need physical robots running around a maze to demonstrate AI.
Seastead this.
If you want to do this right then have your robots spend more time with humans than other robots. This way we can evolve a robot who plays well with other people, than with other robots.
That's what I'm sure my favorite robot SF authors -- Elf Sternberg and D.B. Story -- have planned for their robots. I would love to meet either of their creations.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
"we can win the war by trapping the little fucker in IE6."
Don't the Geneva and Hague Conventions prohibit that level of cruelty?
"This post is an artistic work of fiction and falsehood. Only a fool would take anything posted here as fact."
A simulation I developed around 1987 had 2D robots that duplicated themselves from a sea of parts. They would build themselves up and then cut themselves apart to make two copies. To my knowledge, it was the first 2D simulation of self-replicating robots from a sea of parts. The first time it worked, one robot started canibalizing the other to build itself up again. I had to add a sense of "smell" to stop robots from taking parts from their offspring. As another poster referenced, Philip K. Dick's point on identity in 1953 was very prescient:
http://en.wikipedia.org/wiki/Second_Variety
"Dick said of the story: "My grand theme -- who is human and who only appears (masquerading) as human? -- emerges most fully. Unless we can individually and collectively be certain of the answer to this question, we face what is, in my view, the most serious problem possible. Without answering it adequately, we cannot even be certain of our own selves. I cannot even know myself, let alone you. So I keep working on this theme; to me nothing is as important a question. And the answer comes very hard.""
However, those robots were not evolving. I presented a talk on that simulation at a workshop on AI and Simulation in 1988 in Minnesota, saying how hard easy it was to make robots that were destructive, but how much harder it would be to make them cooperative. A major from DARPA literally patted me on the back and told me to "keep up the good work". To his credit, I'm not sure which aspect (destructive or cooperative) he was talking about working on. :-) But I left that field around that time for several reasons (including concerns about military funding and use of this stuff, but also that it seemed like we knew enough to destroy ourselves with this stuff but not enough to make it something wonderful). At the same workshop someone presented something on a simulation of organisms with neural networks that learned different behaviors. A professor I took a course from at SUNY Stony Brook has done some interesting stuff on evolution and communications with simple organisms: :-)
http://www.stonybrook.edu/philosophy//faculty/pgrim/pgrim_publications.html
Anyway, in the quarter century almost since then, what I have learned is that the greatest challenge of the 21st century is the tools of abundance like self-replicating robots (or nanotech, biotech, nuclear energy, networking, bureaucracy, and others things) in the hands of those still preoccupied with fighting over percieved scarcity, or worse, creating artificial scarcity. What could be more ironic than using nuclear missiles to fight over Earthly oil fields, when the same sorts of techology and organizations could let us build space habitats and big renewable energy complexes (or nuclear power too). What is more ironic than building killer robots to enforce social norms related to forcing people to sell their labor doing repetitive work in order to gain the right to consume, rather than just build robots to do the work? Anyway, it won't be the robots that kill us off. It will be the unexamined irony.
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.