The non-profit, which is funded to the tune of $1 billion, trained a self-learning algorithm to complete a task after a human demonstrated it once in virtual reality.
In this case, the task was stacking coloured blocks.
The team got a programmed robot to reproduce the behaviour shown during the demonstration in the virtual environment.
“We’ve developed and deployed a new algorithm, one-shot imitation learning, allowing a human to communicate how to do a new task by performing it in VR,” OpenAI wrote in a blog post on Tuesday.
The company explained that the algorithm is powered by two neural networks, which are computer systems based on the human brain. One of the neural networks is a vision network and the other is an imitation network.
The vision network is trained with hundreds of thousands of simulated images with different combinations of lighting, textures, and objects, while the imitation network “observes a demonstration, processes it to infer the intent of the task, and then accomplishes the intent starting from another starting configuration.”
While the algorithm was successfully taught how to stack blocks on this occassion, OpenAI said the same technique could be applied to other tasks.
This article was originally published on Business Insider. Copyright 2017.
