Slashdot Mirror
Swiss Experimenter Breeds Swarm Intelligence
destinyland writes "Researchers simulated evolution with multiple generations of food-seeking robots in a new study of artificial swarm intelligence. 'Under some conditions, sophisticated communication evolved,' says one researcher. And in a more recent study, the swarms of bots didn't just evolve cooperative strategies — they also evolved the ability to deceive. ('Forget zombies,' joked one commenter. 'This is the real threat.') 'The study of artificial swarm intelligence provides insight into the nature of intelligence in general, and offers an interesting perspective on the nature of Darwinian selection, competition, and cooperation.' And there's also some cool video of the bots in action."
Of all the Nations, I would never have thought it would be the Swiss who would start the robot apocalypse. I had Germany in my betting pull...
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
To counteract his theories about swarm-intelligence, I sent the researcher link to 4chan.
There are no atheists when recovering from tape backup.
they also evolved the ability to deceive.
Obviously, once you've proved the entity has the ability to deceive, you must distrust any further results.
Do I understand this correctly? On top of superhuman strength and intelligence, we're now making steps toward robot evolution? When robots rule the world, do you think they'll debate whether or not they actually evolved from primitive PCs?
"You fool! We were created in our present form by the great nerd in the sky! Shun the non-believer!"
in this post: http://hardware.slashdot.org/article.pl?sid=09/08/19/185259
I did. Did you understand the meaning behind the paragraphs you cite?
I am confused as to how it was possible to understand the claims it made. Robots don't have genomes and don't eat food. Their "genomes" cannot be recombined.
Robots have code and programmers. What does a random change mean in this context? Do they use faulty dram or mess with the voltage?
Actual robots with flashing lights have a way better chance at going viral on YouTube.
If libertarians are so opposed to effective government, why don't they all move to Somalia?
Real hardware can hold more states than a purely digital system.
I remember reading a paper (can't find it now though - darn it) about a guy who was doing neural net research with Xilinx chips. Same idea. Whenever an algorithm would do well he'd break it into "genomes" and pair them off with other successful programs.
The board was a bank of Xilinx chips, the genomes were the programming files (basically 1s and 0s fed into the configuration matrix), and the goal was to get the thing to turn on and off when you would speak "on" and "off" into a microphone.
It eventually started working. More interesting than that is what happened when he loaded the program into another board. It didn't work.
It turns out the algorithm had evolved to take advantage of the analog properties of the specific chips in that particular board. The algorithm didn't see the board as a digital thing. It saw it as a collection of opamps, amplifiers, and other analog parts. Move the program to a board that is identical digitally, and it failed because the chips weren't analog exact. You wouldn't have seen that behavior in a purely digital simulation.
Weaselmancer
rediculous.
Just with the limited human intelligence, limited resources and limited ability the researchers are able to create great levels of cooperation on mindless robots without any free will. Makes me wonder, if we are designed, as many Intelligent Design advocates claim we are, was the designer "intelligent"? With infinite wisdom and omnipotence and infinite resources, the Designer (or Designers) should have been able to create much more cooperative human beings. No wars. all peace. I wonder how they (the IDists) are able to square their ability ti "infer design" with the obvious "deficiencies of design".
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
I did. Did you understand the meaning behind the paragraphs you cite?
I did! Whenever you hear about a robot/AI "evolving", you should immediately think Genetic Algorithms. Most frequently in association with Neural Networks as is the case here (mentioned lower in TFA).
I am confused as to how it was possible to understand the claims it made. Robots don't have genomes and don't eat food. Their "genomes" cannot be recombined. Robots have code and programmers. What does a random change mean in this context? Do they use faulty dram or mess with the voltage?
Robots have code. That code can be (is) represented as a series of bits. That series of 1s and 0s can be considered the "genome" of the robot. And can you randomly change bits, or recombine portions of bit patterns? Absolutely. When you have a "population" of such bit patterns which you test out, then take the best ones and copy them, randomly mutate them, and recombine them together, then test the new population and repeat, you have a Genetic Algorithm.
Now that's possible but time consuming to do with the actual instruction bytes of the AI. In the case of Neural Networks with Genetic Algorithms, the "genome" is actually just the organization of the neural net -- the connections and weights between the neurons.
The robots still have programmers, but the programmer is not directly writing instructions for the AI to follow. Instead, they're writting an interface between the robot's sensors and the neural network, the neural network code itself (not its organization), and the code that does the genome mutation/recombination. From there, the AIs "evolve" and "learn" on their own and arrive at often fascinating solutions to problems.
For example in this case, the programmers did not teach the robots how to deceive each other. That behavior emerged on its own from the genetic algorithm. So, there really isn't any hyperbole at all in the summary/TFA. Though it would be helpful if they at least mentioned genetic algorithms so skeptical people have something to google. :)
The enemies of Democracy are
One question that intrigues me is just how human-readable the code produced by such genetic algorithms is. Some of the practical promise of this work is that it produces problem-solving code in ways very difficult from that of human programmers -- but how can such code be maintained by humans? It's a bit like making an engineer try to figure out how your lower intestine works.
I imagine that there might be interesting results that come from putting objects into an environment where you don't control all the variables. I've heard of cases where the robots end up using features of their own hardware (which is generally cobbled together from off the shelf parts) that the researchers never anticipated.
it sounds like you are now accepting that the article is misleading.
You're agreeing that the terms are not what a normal reader would construe them to mean.
The terms are a (very good) metaphor, and the article is not at all misleading. I would have thought this would be obvious.
if the experiment wants to show anything, the methodology has to be more transparent so that we can know whether to consider its "genome" as really a genome or are something more banal.
The entire point of this sort of research is that the "genome" in the bots is analogous to, but far simpler than, a biological genome, and the means of selecting which "genomes" to generate the next "generation" from is analogous to how genomes are selected in biology (either "natural selection" like you find in nature or "artificial selection" like you get with farmed crops or dog breeding).
In what way is it not transparent?
if the semi-random is really just someone going through and changing parameters in a config file (or using a script to do it), then it's not really random at all.
Believe it or not, computers actually can generate effectively random numbers.
Because intelligence isn't just a software thing. At least not in humans.
I recall reading about field programmable gate arrays being used in an experiment with genetic algorithms. They wanted to force the FPGAs to evolve to tell the difference between two different frequency sounds. Eventually they wound up with chips that accomplished the task in a variety of ways - ways that worked but for no explicable reason, some of them being ways that took advantage of tiny differences in the individual (identical, at least from a manufacturing perspective) chips, and even that required slight differences in the room's environment. This was years ago.
Simulations won't have those little idiosyncraces between individual units and thus might miss a huge component. Variation among individuals that is only in software misses the whole concept of variation between individuals that comes about from hardware, and also from the interaction between the two.
Since I can't tell them apart, I treat all ACs as the same person.
Same story more than a year ago: http://hardware.slashdot.org/hardware/08/01/19/0258214.shtml
/.? Really?
And offtopic: $&^@%! Taco, what's up with the popups that sneak past Firefox popup blocks? I've dutifully allowed advertising to continue, despite having that checkbox I could click to turn ads off for good behavior. Do I really need to turn on adblock and noscript for