Imagine this: you've become the slightest bit parched and you have an empty glass. There's a full bottle of beverage just waiting to be poured into your cup — but it's so far. Besides, pouring is a servant's job, anyway. Thankfully, Cornell University has recently furnished you with a robot servant. And while robots of one sort or another have been around since the 1980s, this one is new.
You needn't call out to your new robot like some sort of commoner. No, no — just reach out for the bottle and your robot butler will predict your desire for more liquid in your glass. Then, like a good servant, it will scoot on over, pour your drink for you, and then stare blankly at you, awaiting your next whim.
Such is the vision of Cornell's Ashutosh Saxena, assistant professor of computer science. To accomplish this dream, Saxena and his team have armed a humanoid robot with a Kinect 3D camera and a library of 3D videos that exemplify common human habits. Habits like drinking coffee, for instance.
The idea is that the robot will record further videos of you completing tasks about the house and build a "vocabulary" of actions. Once this vocabulary is broad enough, the robot will be able to complete more complex tasks on its own.
Currently, the robot is able — with just under 85 percent accuracy — to predict your actions, three to ten seconds before you actually take them. The one visible drawback is that, though the robot can see what you want to do before you do it, it can't currently save you any time by doing things for you. For that you'd need a much more spry robot. Check out the very cool anticipating robot, along with its somewhat lacking reaction time, in the video below.