Home → News → Optical Illusions Fool Computers Into Seeing Things → Full Text

Optical Illusions Fool Computers Into Seeing Things

By New Scientist

December 15, 2014

[article image]


University of Wyoming researchers are studying a particular type of image-recognition algorithm called a deep neural network (DNN), combined with a second algorithm designed to evolve different pictures. The algorithms, working in conjunction with human judgment, have previously created images of apples and faces, and the researchers wondered if replacing the human with a DNN, to work alongside the genetic algorithm, would work as well, resulting in a program that could generate creative pictures by itself. "We were expecting that we would get the same thing, a lot of very high-quality recognizable images," says University of Wyoming researcher Jeff Clune. "Instead, we got these rather bizarre images: a cheetah that looks nothing like a cheetah."

The researchers used AlexNet, a DNN created by University of Toronto researchers in 2012. The researchers found the genetic algorithm produced images of seemingly random static, which AlexNet declared to be pictures of a variety of animals with more than 99-percent certainty. The algorithm's confusion is due to differences in how it sees the world compared with humans, according to Clune. "All optical illusions are kind of hacking the human visual system, in the same sense that our paper is hacking the DNN visual system to fool it into seeing something that isn't there," Clune says.

From New Scientist
View Full Article

 

Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA

0 Comments

No entries found