Robots Learn to Lie and Deceive Each Other in Search for 'Food'

Technology Review reports that a team of scientists at Switzerland's Ecole Polytechnique Fédérale de Lausanne are soon to release a study on robots equipped with artificial neural networks and programmed to locate 'food.' When a robot neared the 'food,' it flashed a blue light so other robots could also find it. With limited space around the 'food,' the robots soon learned this wasn't the best idea. The researchers copied and combined the artificial neural networks of the most 'intelligent' robots, and made a few changes to the code to mimic biological mutations. As a result, the robots 'evolved' -- learning not to alert other each other to the food. After a few hundred (increasingly intelligent) 'generations,' the majority of robots didn't flash a light at all.
The researchers say this experiment will help scientists better grasp how biological communication systems evolve. In this case, it might not have happened very quickly, but the results, which will be published in the Proceedings of the National Academy of Sciences, are pretty amazing. Pretty soon, we'll have those pesky robots to deal with at the dinner table, too. There goes any shot at seconds on Thanksgiving. [From: Technology Review]





Whitney Houston Autopsy: Cause of Death Determined?
Whitney Houston, Bobbi Kristina: Late Singer's Daughter Hospitalized
Adele Five-Year Break? Singer Plans to Focus on Relationship, Write 'Happy Record'
Jennifer Hudson Whitney Tribute: Grammy President Reveals Why Singer Was Chosen for Musical Memorial
Grammy 2012 Winners' List: Adele Sweeps Music's Biggest Night
3 Economic Misconceptions That Need to Die
5-Hour Energy: A Success Equal Parts Caffeine, Chemistry and Meditation
People With Easy-To-Pronounce Names More Likely To Succeed, Study Says
Katy Perry Grammy Performance 2012: Did the Diva Diss Her Ex-Hubby With Revealing New Song?
Whitney Houston Dead: Stars React to Legend's Sudden Death














Comments
4
Subscribe to commentsRandyAug 20th 2009 4:47PM
can they say "I'll be back"?
KJAug 20th 2009 5:51PM
The description of the experiment is pretty brief, but this seems pretty manipulative. You create a situation where robots that flash lights get less "food" and you assume that getting less "food" means your traits won't be passed down. Then you change the code to reward what you judge got more "food," and because of that, the robots evolve. This would be a more interesting experiment if the code could somehow pass itself down to generations and it did so based on success at getting food, as opposed to some programmer watching what happens and adjusting the code himself. Even that wouldn't take into account factors beyond gathering food, for example, if the other robots ostracize liars.
flgunnerAug 20th 2009 6:07PM
Is this so they can find the food before we do?
What a waste of money. Is there truly any useful logical reason for "studying" how these robots react? Unless they are being trained to take from humans by whatever means.
Davey RockefelerAug 20th 2009 10:06PM
Most computer scientists think within say about 70 years human beings could be at risk as a species from intelligent machines. Why would they want us around? Maybe a few controlled support people? Computers run our lives now anyway and we can't pull the plug. I used to sit and have a scope probe and chase their bits around trying to keep them up and running but now-people are almost out of the loop. Sooner or later.......Going back to my planet now.