Google’s deep learning software teaches itself how to play computer games without prior knowledge of the controls, the goals or the methods of game play. You’re sure machines will never be intelligent enough to replace you? Think again.
This crazy-ass video is of the Hubo Humanoid Robot, winner of the 2015 Darpa robotics challenge. The robots are given goals. They have to then make decisions about how to best accomplish each one. It’s incredible to perceive the scope of tasks that this guy can handle, from driving a car to turning a valve to utilising a yellow power tool to slice through a wall.
The win proved a bit of a controversy because Hubo elected to push the obstacles aside instead of clambering over the rubble. It has been suggested that this action was cheating, but what’s more mind-blowing is a robotic mind having the capacity to seek out a loop hole, to bend rules in order to find the simplest solution. Ruminate on that for a second or two.