Robots Evolve, Learn to Become Better Predators
There are some interesting robotic evolution experiments going on over in Switzerland that could have horrifying results.
Researchers from the Laboratory of Intelligent Systems of the Ecole Polytechnique Fédérale of Lausanne, Switzerland have been experimenting with evolutionary robotics, and achieved successful results that should be alarming to the human race. Some of the researchers' robots are learning to cooperate with each other, while others are learning to become better predators.
The experiments are based on Alan Turing's theory that "intelligent machines capable of adaptation and learning would be too difficult to conceive by a human designer and could instead be obtained by using an evolutionary process with mutations and selective reproduction." The lab's robots were "controlled by a simple neural network, which mutated randomly." Each successive generation would improve based on the last generation's success or failure with a certain task.
One of the experiments involved a co-evolutionary process between both predator and prey robots. The prey robots were twice as fast as the predator robots, but predators could detect prey from a further distance. Eventually, the predator robots learned to approach prey from the sides, where there were no sensors. The prey counteracted this strategy by rotating quickly in place, and backing away from predator robots with sensors facing them. When the prey evolved to move at maximum speed along the experiment area's walls, the predators evolved to lay in wait like a spider.
These experiments also created robots that would cooperate with each other towards a common goal. In this case, it was pushing different sizes of tokens, with one size that couldn't be pushed without multiple robots. The larger token gave the robots a greater amount of "points," or evolutionary value. In the experiments where cooperation was required, the robots learned to succeed every single time.
I don't think I have to tell anyone how bad this is for humanity's future. Robots will someday become predators of the human race, constantly get better at it no matter what we do, and will also learn to cooperate at enslaving us. This is why I'm starting H.A.R.M (Humans Against Robot Masters). We accept anyone, but you have to kill an evil robot to get in. Down with metal, up with flesh!
Source: PopSci [http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1000292]
Permalink
There are some interesting robotic evolution experiments going on over in Switzerland that could have horrifying results.
Researchers from the Laboratory of Intelligent Systems of the Ecole Polytechnique Fédérale of Lausanne, Switzerland have been experimenting with evolutionary robotics, and achieved successful results that should be alarming to the human race. Some of the researchers' robots are learning to cooperate with each other, while others are learning to become better predators.
The experiments are based on Alan Turing's theory that "intelligent machines capable of adaptation and learning would be too difficult to conceive by a human designer and could instead be obtained by using an evolutionary process with mutations and selective reproduction." The lab's robots were "controlled by a simple neural network, which mutated randomly." Each successive generation would improve based on the last generation's success or failure with a certain task.
One of the experiments involved a co-evolutionary process between both predator and prey robots. The prey robots were twice as fast as the predator robots, but predators could detect prey from a further distance. Eventually, the predator robots learned to approach prey from the sides, where there were no sensors. The prey counteracted this strategy by rotating quickly in place, and backing away from predator robots with sensors facing them. When the prey evolved to move at maximum speed along the experiment area's walls, the predators evolved to lay in wait like a spider.
These experiments also created robots that would cooperate with each other towards a common goal. In this case, it was pushing different sizes of tokens, with one size that couldn't be pushed without multiple robots. The larger token gave the robots a greater amount of "points," or evolutionary value. In the experiments where cooperation was required, the robots learned to succeed every single time.
I don't think I have to tell anyone how bad this is for humanity's future. Robots will someday become predators of the human race, constantly get better at it no matter what we do, and will also learn to cooperate at enslaving us. This is why I'm starting H.A.R.M (Humans Against Robot Masters). We accept anyone, but you have to kill an evil robot to get in. Down with metal, up with flesh!
Source: PopSci [http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1000292]
Permalink
