HOW TO TEACH ROBOTS NEW TRICKSDate: 2018-05-23
Another week, and we are another step closer to a com/article/the-reality-of-an-artificial-intelligence-apocalypse" target="_blank">robot revolution.
com/article/the-reality-of-an-artificial-intelligence-apocalypse" target="_blank">robot revolution.
Researchers at NVIDIA are teaching robots to complete tasks by simply observing human actions. The team, led by Stan Birchfield and Jonathan Tremblay, developed a deep learning system "to enhance communication between humans and robots".
Using NVIDIA Titan X GPUs, the research team trained a series of neural networks to execute actions based on a single real-world demonstration.
“In order for robots to perform useful tasks in real-world settings, it must be easy to communicate the task to the robot; this includes both the desired end result and any hints as to the best means to achieve that result,” a recently published paper said.
Live video of a scene – someone stacking coloured cubes, for instance – is fed into the neural network, which infers the positions and relationship of objects; another machine then generates a plan to recreate those perceptions.
The robot automatically picks up its own mistakes, if it messes up at any stage of the process, it realises it hasn't achieved the goal yet and tries again. However, before it does anything, the robot also produces a description of its plan designed to be readable by humans, so its supervisors can check that it hasn't misinterpreted the task, and re-teach it if necessary.
Finally, an execution network reads that proposal and generates an intelligible description of steps, which a user can edit directly before the android even moves.
The robot can be seen in action in the video below.