Columbia Engineers Create Robot that Predicts Behavior Of Other Robots

Scientists won’t stop until Judgment Day is finally here.

Up next on our path to obliteration is this research out of Columbia University, where engineers are busy exploring what they call the “Robot Theory of Mind.” What they have so far is pretty interesting.

In short, they put together a little experiment involving two robots, one an observer and the other an actor. The goal of the Observer Robot was to visually predict the trajectory, or behavior, of the Actor Robot as it sought out the closest of two green circles.

Here’s the trick: An obstacle would be placed near one of the circles, preventing the Actor Robot from seeing it. This would cause the actor to either move toward the green circle it could see, or not at all. But the Observer Robot, high up on its perch, could see everything.

From there, the researchers trained the observer to understand the actor’s behavior and lack of visibility:

“After observing its partner puttering around for two hours, the observing robot began to anticipate its partner’s goal and path. The observing robot was eventually able to predict its partner’s goal and path 98 out of 100 times, across varying situations — without being told explicitly about the partner’s visibility handicap.”

You might call it a very rudimentary form of empathy.

The researchers believe that visual behavior modeling may be key in creating robots with social cognition. But they also provide a slight warning about what this could mean in the future.

As study co-author and Mechanical Engineering Professor Hod Lipson said, “We recognize that robots aren’t going to remain passive instruction-following machines for long…”

Okay, then.

The study was conducted by Boyuan Chen, Carl Vondrick and Hod Lipson at Columnia Engineering’s Creative Machines Lab, and you can read their full paper right over here.


Rob Schwarz

Writer, blogger, and part-time peddler of mysterious tales. Editor-in-chief of Stranger Dimensions. Might be a robot.

Read More