'Robo Brain', a collaborative project between two Indian-origin researchers, is a large-scale computational system that comprehends everything from publicly available Internet resources, in order to teach robots natural human behavior. This complete data-bank for robots shall help the machines with finding keys, pouring a drink, putting away dishes and not interrupting two people when they're amidst a discussion. Lead researcher Ashutosh Saxena, assistant professor of computer science at Cornell University (New York) explained about the project, stating that out computers and cell phones have access to all information we require, and if in case a robot encounters a situation it has not experienced before, it can simply query Robo Brain in the cloud.
At present, as you're reading this article, Robo Brain is downloading and processing about 1 billion images, about 120,000 YouTube videos and nearly 100 million how-to documents and appliance manuals. All this information is being translated and saved in a robot-friendly format that the robots will understand when required, Saxena stated. As per him, Robo Brain shall process images in order to pick out the objects in them and by connecting these images and videos with text shall it identify objects and their usage, coupled with human language and behavior.
Take for example, of a robot comes across a coffee mug, it can learn from Robo Brain not only that it's a coffee mug but also the liquid content that can be poured in or out of it, that the mug could be grabbed via the handle, and that it must always be held upright when it's full. Aditya Jami, the large-scale database designer of Robo Brain, states that it'll look like a gigantic branching graph "with abilities for multi-dimensional queries."
Similar to a human learner, Robo Brain shall have teachers thanks to crowdsourcing, concluded Saxena who presented the project at 2014 Robotics : Science and Systems Conference held in Berkeley lately.