Last time when I was in Tokyo my friends were telling me about their domestic robot projects, they were very cool ideas and didn’t sound too far away. I was half-jokingly talking about how I see that as another platform play. The problem with machines is their limited ability to learn, unlike humans. Today robots are a long way from matching human emotional complexity, Japan is working hard to create this future — where humans and intelligent robots routinely live side by side and interact socially and won’t get into any argument or a fight. I guess they will program them to sing Karaoke if they ever get too upset by humans (at least for the Japanese domestic version). Other export version will be programmed to dance or act silly to make human calm down.
Robots have been part of Japanese factories for many years, but they rarely leave the factory doors. Robots can do other things, they can make sushi, plant rice and tend paddies, serve as bartenders, work in gas stations, vacuum your living room, massage your back and serve green tea. And then next they will evolve into surrogates? Yes, I've watched that movie on the plane.
A team of Swiss engineers have applied Darwinian selection to robot development, producing robots that can walk, cooperate and even hunt each other. They were playing with the idea that robots that are controlled by simple neural network which can mutate randomly. This is not Sci-Fi. The input neurons of the neural network will be activated by the robot's sensors and the output neurons will control its motors.
Each robot had a different 'genome', describing different connections between neurons. This resulted in unique behavior and fitness - how fast and straight it moved, for example, or how often it collided with obstacles. Darwinian selection was then imitated, by choosing the genomes of the robots with the highest fitness to produce the next generation.
Machine based competition for learning, that’s what it is. Genomes were paired, and random mutations such as character substitution, insertion, deletion, or duplication were applied. One of the discoveries was within 100 generations, robots were able to move without collisions in a maze. I can see our next generation of vehicles with this technology, no more car accidents and drink and drive will not cause any problems, that is you wont’ press the wrong button.
Their experiments also discovered that groups of unrelated robots - those with randomly differing genomes - invariably took the selfish approach and went for the small tokens. But those with similar genomes generally pushed the larger tokens, cooperating to raise the fitness of the whole group - and thus reducing their own chances of 'winning'. So game theory for robots, can’t get better than this.
The research was done by Dario Floreano and Laurent Keller from Laboratory of Intelligent Systems, Ecole Polytechnique Fédérale de Lausanne and Department of Ecology and Evolution, University of Lausanne, Switzerland.
