Robot evolution

Harpo

Getting away with it
Joined
Sep 23, 2006
Messages
3,727
Location
The edge of the world. Yes, really.
80. Robots Evolve And Learn How to Lie | Robots | DISCOVER Magazine

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.”
 
Fascinating stuff. I'm not very well informed about artificial intelligence, but find it very interesting. Thanks for posting this:)
 
Fascinating indeed. How many generations are needed to surpass human intelligence? Is 50000 a reasonable number?
 
This reminds me of an early tale called, IIRC, 'Callahan and the Wheelies'...
 
Why would we want a robot that can lie?

I mean.............think about it..............
 
Why would we want a robot that can lie?

I mean.............think about it..............
Who would have the resources to go through sufficient generations with enough complexity?

Governments, big business, the military – all paragons of truthfulness who could never consider a distortion of reality a positive factor, even if it gave them an advantage.

But evolution doesn't give you what you want, unless your limit conditions are very well defined; it gives you what can best survive in a given set of conditions. Intelligent design might be a bit better, but that's what's being used to get the computing we have now, and enough random factors mix in to keep me confused even so.

I am reminded of Rudy Rucker's "boppers". (Software, et al)
 
Why would we want a robot that can lie?

I mean.............think about it..............

If you've seen 'the invention of lying' then surely there are a thousand reasons that we would want a Robot that could be lenient with the truth. Lying has real evolutionary advantages. Just look at women. ;)
 
Women use subtlety and subterfuge, its not our fault we think on a higher evolutionary scale then the easily persuaded male.

:p

Robots are machines. They should not evolve.
 
So, how many lying robots does it take to change a light bulb?
 
Fascinating stuff. I'm not very well informed about artificial intelligence, but find it very interesting. Thanks for posting this:)

There is a distinction between "AI" and Robotics. They aren't necessarily always combined, but they can be. In the strictest sense, Artificial Intelligence is defined in the software only; as it is the only software that can re-write itself. This can be useful, and ultimately will be what contributes to the "evolution" of robots. It is also dangerous. If the programmer isn't fully cognizant of every possible outcome (which is an enormously burdensome task) the robot may begin to display unpredictable behavior. This is, of course the notion that is exploited in the "Terminator" movies. A fail-safe switch or power cutoff is always employed in these type of robotic experiments.

A safer, and easier concept to implement is one that's been around quite awhile in electronics and software. It's called "control theory, or "feedback". The robot can add "memories" of the consequences of their actions, and learn what to do, and what not to do. They would of course have a basic list of do's and dont's programmed in. This is the less strict form of AI, where there is software code added, but none of the original software is re-written "on the fly".

That last phrase is the perfect lead-in to my next suggestion, that Jet airplanes are essential big powerful robots. With all of the advances in inertial navigation, and GPS positioning, Passenger Jets can take off, stay on course, and land. The tricky bit about lining up with the passenger loading ramp could be done with sensors and such, but it's just too expensive. That is always the anathema of such advancements. Cheaper isn't necessarily better.
 


I agree with you on this one, Moonbat, and to answer the question, they should "evolve" in every sense of the word because, although the ARE artificial, they are INTELLIGENT artifice.


Just because something's natural doesn't mean it should evolve-do rocks evolve?

Just because something is artifice doesn't mean it shouldn't. Intelligent design evolves to adapt and survive. Even plants have intelligence to some certain extent. Rocks and water doesn't, just like robots have intelligence where a monkey wrench doesn't.
 
Evolution doesn't require intelligence; the greater part of evolution took place before intelligence was even an option. Evolution requires passing of characteristics, so a successful design has more chance of being copied than a less successful one (in many ways, this could be the definition of 'success'. A supremely effective predator with no sex drive would, in the evolutionary stakes, be considered as a failure). In many ways, an outside lifeform being the chief selection apparatus (as in, say, bees and flowering plants developing in parallel) is a standard mechanism, so the monkey wrench in question could be said to have evolved from its flint hand-axe ancestors due to continuous human intervention. (I wouldn't, though. For me, evolution demands random, rather than planned change – and human evolution is blocked due to intelligence. Still, the idea of the next step in human evolution being calculated, rather than leaving the direction open to survival tests (for the time being, being lazy, ignorant and insensitive leads to more offspring than its inverse; it is fortunate that evolution takes such a long time in a genetically diverse species to cause any noticeable changes.)

So, robots can be designed for random diversity, and desired tendencies selected for, as in breeding poodles; or they can be designed to hand on bits of their software, producing their own future generations, and given insufficient resources, so survival means something. Either of these situations can be seen as a form of evolution; if we can breed both a Great Dane and a Yorkshire Terrier out of primeval wolf or jackal stock (without speciating, I add) might we not leave imprecision in our blueprints, and evolve true intelligence from mere rapid calculation?
 
It doesn't require intelligence, per se, but it does tend to require some form of consciousness and "life", is what I meant.

I know what you mean Karn, but I'd say tread lightly. The eventual implication here is that man can create artificial life that is indistinguishable from human life, and the conservative Anti-Science types might have a fit about whether you could call a robot's existence "Life."
 
There's a bit of muddy water in this discussion.

First, on the subject of generations, there isn't in fact much of a correspondence with the biological concept of generations. Robots with the ability to build their own offspring would be free to build them any time, and their offspring could be immediately endowed with all the learning of the parent. Thus, in a sense, there is only ever one generation, though there may be many individuals in it. On the other hand, evolution of AI itself takes place at a software level (as has been correctly pointed out), and here 50,000 `generations' can actually be computed in a fraction of a second. In effect, robots would be very quick to adapt to changing circumstances, whereas it will take humans a century or two to get their heads around global warming....

Even though AIs can adapt quickly to circumstances, for robots to be considered alive and form part of society, they have to go through the same sort of learning processes that humans go through; they must make the mistakes that children make, and experience the stages of growing up that adult humans have experienced. They must be schooled in the same way (and on the same time scale!) that humans are schooled. Ultimately a `living' robot will be the sum of its experiences in the real (not mathematical!) world. I think the danger that robots might then realize that they can form a powerful clique in our society and work to decimate the human race is very real; ultimately it is observed that physically strong people exert themselves over others in all walks of life, and there is no reason to believe that robots won't take that attitude to an extreme. :eek:

ps. I'm a fan of technology, and in a way I kind of hope it happens... is this not the next big step in our own evolution? To put our accumulated wisdom into living machines that live forever and travel between the stars, explore the galaxy and assimilate new life-forms to ensure continued domination? Resistance is futile!
 
Last edited:
The idea of a robot that evolves is terrifying.

It is a distinct possibility. It would require some means of storing information that can be altered from generation to generation. If each robot carried an internal memory with instructions for building its duplicates, and if that memory could change in some random fashion, then many generations later, if humans did not intervene, we could end up with a generation of robots much more evolved.

There was an SF story I read many years ago. Forgot the title and author. Some marine mining automated solar powered barges were left on a planet to mine resources and to occasionally build more barges. Humans did not return for tens of thousands of years. When they got back, they found an entire global ecosystem of robot descendents - thousands of species of mechanical life forms. Some collected light energy. Some preyed on the energy collectors. Some were stupid. Some were intelligent.

The conclusion to this story was when the humans decided that the planet now belonged to the evolved robots, and they left. However, there is no guarantee that evolved robots would ignore humans, or be allies. They could see humans as competitors. And then the trouble starts.....
 

Similar threads


Back
Top