A while back we talked about a group of scientist that taught robots how to share, our main concern with that was, “WHY!? all were doing is teaching them how to share our body parts for energy! when we have to fight them” well, now some Smart Masses up in Switzerland developed a robot that can learn algorithms that allow robots to learn from human mistakes…
So pretty much all of us Oxygen breathing earthlings are making a robot that will learn how to kill us after they learn that were just a bunch bothersome plague-ridden planet earth annoyances… Similar to Skynet on the Terminator series.
The way the robot works is, a Human guides the robot through a flawed attempt at performing a specific task. Watch the video below to see what they have them do, for one example they have a small block, the purpose of the robot is to make the block stand on it’s side by bumping it from the bottom. The Robot performs two flawed attempts – while it’s being controlled by a human – the robot then learns from the flawed attempts and using an algorithm the robot calculates what went wrong in the human guided example. After it performs a number of calculations, it finds a way to succeed via Trial and Error.
Once the robot figures out how to do it, it instantly memorizes that and it can perform the operation in the future. Now we all know what happened in Terminator, or iRobot when the robots learned that we were all failures, or even Space Odyssey with HAL 9000, because of that, the scientist are only using the algorithm on a single hyper-learning appendage. If this thing ever breaks lose, and learns how to control the internet, I bet we can all agree… we as humans would fail at shutting it down, and instead, that robot would learn to defend it self like the movie eagle eye.