An Elon Musk-backed artificial intelligence research firm called OpenAI has created a way for
A blog post on Tuesday by OpenAI outlined how
An OpenAI researcher put on a VR headset and imitated picking up the blocks. Two so-called neural networks – a computer system that mimics the human brain – then process this allowing a robot to carry out the task.
The vision network uses the robot's camera to determine the position of the blocks relative to the robot. It is trained by being fed thousands of simulated images showing blocks in different positions with various lighting and textures. This way, the network is able to recognize an object in the physical world even though it's never seen an image of the block with a real camera.
The second neural network
"Imitation allows humans to learn new behaviors rapidly. We would like our robots to be able to learn this way too" Josh Tobin, a member of the technical staff at OpenAI, said in a video posted on the organization's website.
"With a single demonstration of the task, we can replicate it in a number of different initial conditions. Teaching the robot how to build a different block arrangement requires only a single additional demonstration. "
OpenAI is a non-profit that has got around $1 billion in funding and is backed by Tesla boss Elon Musk, Peter Thiel and LinkedIn co-founder Reid Hoffman, among others. The group states that its mission is to "build safe AI, and ensure AI's benefits are as widely and evenly distributed as possible." One of its goals is to build a household robot.
The latest development, which is seen as a big step for general purpose robots, could lay the groundwork for this. Even though the robot was stacking blocks, OpenAI explained that this could apply to other tasks.
"Nothing in our technique is specific to blocks. This system is an early prototype that will form the backbone of the general purpose robotic systems we are developing here at OpenAI," Tobin said.