Researchers at MIT have given robots everything they need to take over the world. A new algorithm called Estimation, Extrapolation, and Placement (EES) allows robots to train themselves without human help. While this robot self-training algorithm may seem convenient, it's sure to make quite a stir among robot pessimists.
In addition to enabling robots to train themselves, EES algorithms enable them to identify weaknesses in their skills. The robot can then use those weaknesses to determine where it needs to improve its skills. The robot does this by using its vision system to assess its surroundings and the task it is given, such as cleaning or sweeping a room.
Unitree Robotics G1 humanoid robot cooking breakfast? Image courtesy of Unitree
The robot can use the EES to determine if additional practice is needed to improve its performance. If so, the robot's self-training algorithm creates and uses training material for the robot. The researchers tested the algorithm on one of Boston Dynamic's Spot robotic dogs, which is already very good at simple tasks. However, this time Spot got the job done smarter.
Of course, there are many concerns about the current state of robots and AI. Many worry that AI will overthrow humanity, and even the godfather of AI believes that this could happen someday. But it's also clear that the folks at MIT likely designed this algorithm to make robots better at simple tasks, not robots that will wage war or do other dangerous acts.
Technology. Entertainment. Science. Your inbox.
Sign up to receive the most interesting tech and entertainment news.
By signing up, you agree to our Terms of Use and acknowledge our Privacy Notice.
But that's unlikely to stop others from using this or similar systems for specific applications. We can only hope that China doesn't get hold of this robot self-training algorithm for its rifle-toting robot dogs. The EES algorithm is featured in a new paper published on the preprint server arXiv.
