By James Riordan

At the 29th annual conference of the Association for the Advancement of Artificial Intelligence (http://www.aaai.org/home.html) researchers presented a paper indicating that they have discovered a new method to teach robots how to use tools – by watching videos on YouTube.  The researchers, from the University of Maryland and the Australian research center NICTA, employed the latest use of a type of artificial intelligence called deep learning.

A hot topic in scientific circles as of late, deep learning involves using training systems called artificial neural networks to absorb and process large quantities of information derived from audio, images, and other inputs.  The AI system not only stores the information, presents it in a new manner including making and receiving inferences about the information as a response.

Using this method, robots can be taught complicated tasks such as cooking in the same way humans can learn them – by watching a video on Youtube.  The researchers employed convolutional neural networks, similar to the type now in use in social media programs such as Facebook and other companies, to identify the way a hand is grasping an item, and to recognize specific objects. The system also predicts the action involving the object and the hand.  To train their model, researchers selected data from 88 YouTube videos of people cooking. From there, the researchers generated commands that a robot could then execute.

“We believe this preliminary integrated system raises hope towards a fully intelligent robot for manipulation tasks that can automatically enrich its own knowledge resource by “watching” recordings from the World Wide Web,” the researchers concluded.

Read the complete report, “Robot Learning Manipulation Action Plans by ‘Watching’ Unconstrained Videos from the World Wide Web.”.

___