Precognitive robot learns to predict our future needs

Remember those days when we stood a chance against the marching robot legions because we could still devise a plan that 'the robots would never think of'? Me too.

They're gone now, of course, as Ashutosh Saxena and his team in the Cornell Department of Computer Sciences have a robot that can actually tell what you're going to do in the future.

Did you just pick up a casserole dish to put into the fridge? The robot will watch you pick it up, run algorithms to predict if you're going to put the dish into the sink, on the counter or in the fridge, and then based on the position of your body (and what information it's gathered from you in the past), it will pick out the most likely one and help you with it. In this case, it will roll forward and pull the fridge door open for you.

[ Related: Secretive robot space plane mission passes 5-month mark ]

As the video shows, it can even tell when it needs to hold off on assisting you, so as not to waste precious beer.

This project builds on work the team has been developed over the last few years. The first step was to design a system of 'scene understanding', where a robot can recognize objects in its view and assign different labels to them. It even gives vertical and horizontal surfaces different labels, to tell the desk top from the desk sides. That way, it doesn't interpret the desk side as a flat surface and try to put something down on it. The next step was to give the robot a way to recognize human activities, which it does by mapping a stick figure over any person in sight, and then tracking the stick figure as the person moves.

From there, it was a matter of programming a way to predict the future.

I've worked with computer models like this, but they didn't track humans or objects. They tracked air pollution, by mapping out the winds from various weather forecasting models, and then dropping an individual particle into the wind streams to watch where it goes. With different models giving different directions the wind can blow the particle, computer algorithms overlay all the possibilities and figure out the most likely path the particle will take.

This research follows the same idea, but instead of using wind fields, it uses the realm of possible actions that the person can take, and apparently it's easier to predict what we're going to do than to tell exactly which way the wind is going to blow.

[ More Geekquinox: Massive asteroid makes closest pass by Earth on Friday ]

So, it probably goes without saying that this is an amazing accomplishment for the researchers, but on the down-side, this does mean that there's another arrow gone from our quiver of possible ways to defeat the robots when they eventually rise up against us. We've probably already lost the ability to make them overload their circuits by giving them complicated problems to solve, and it likely won't be long before we can't use our emotions to confuse them either. In fact, it's already been shown that we may fall prey to them using our own emotions against us!

I guess as robots become more advanced, we're just going to have to find a way to keep them from rising up. There's no guarantee that Asimov's Laws will protect us, so we might want to move forwards on giving them the same kinds of legal rights we enjoy. After all, nothing helps promote peace better than equality.

Geek out with the latest in science and weather.
Follow @ygeekquinox on Twitter!