Robot Predator Hunts Prey with Digital Retina

Swiss researchers crossing neural network A.I. technology with predator-prey dynamics.

Important update from the impending robot revolution: Scientists in Switzerland are teaching artificial intelligence systems how to behave like predators and hunt down their prey.

The annals of science fiction are full of cautionary tales in this regard, of course, but the Swiss researchers insist they have only the best of intentions. (Isn't that what they all say?)

RELATED: Realistic Robot Lady Cheerfully Agrees To Destroy Humans

But seriously, folks: According to a report over at Motherboard that's quickly making the rounds, scientists at the University of Zurich's Institute of Neuromatics are using specialized cameras and neural network technology to program sophisticated predator 'bots. The idea is to use predator-prey models that encourage robots to efficiently scan their environment, find targets, and stalk them follow them.

WATCH VIDEO: What If the President Were a Robot?

In terms of practical applications, the researchers hope the find-and-follow system will result in improved models for self-driving cars or autonomous drones. But researchers say there could be other applications, too -- like luggage or shopping carts that follow you around.

The key to the whole initiative is something termed a "silicon retina." Predators need sharp vision to follow their prey in high-speed situations and traditional cameras aren't quite up to the task. The silicon retina -- part of the larger EU-funded VISUALISE project -- mimics the organic functions of the eye.

RELATED: Bullied and Shoved Humanoid Robot Stays Upright

Rather than analyzing moving images frame by frame, the digital retina processes information pixel by pixel, assessing the movement of objects with greater accuracy. By instantly crunching the numbers on trajectory or relative illumination, the digital retina can better track path of movement, especially with fast-moving objects.

The retina feeds this information to the robot's brain, a deep learning neural network that can process vast amounts of data instantly. With repetition and time, the robot essentially teaches itself how best to track and follow objects.

The Swiss researchers have been testing robots using the system to track other human controlled robots, with appreciable success, it seems. One final note: Researchers say 'bots using the predatory-prey system definitely get better with practice. That's a nice detail to sign off with, I think.