Carnegie Mellon's computer program known as NEIL (for Never Ending Image Learner) is designed to teach computers "common sense" by continuously analyzing internet images, futurity.org reported. The idea is to establish nuance and context, for one doesn't learn what an object is and how it works from a single picture or experience, but from a lifetime of exposure.

For example, if NEIL analyzes enough images of the word "apple," it should be able to distinguish Apple, Inc. from the fruit, the situations in which one is more appropriate than the other, and the relative popularity of each. Every photo is a learning experience for Carnegie Mellon's baby, NEIL, which "craws, sees, and learns," according to its website.

"Images are the best way to learn visual properties," said Abhinav Gupta, assistant research professor in Carnegie Mellon University's Robotics Institute. "Images also include a lot of common sense information about the world. People learn this by themselves and, with NEIL, we hope that computers will do so as well."

NEIL launched in July and has sifted through over three million images during its nonstop run, according to futurity.org. The software has categorized 1,500 types of objects and 1,200 types of scenes.

NEIL also subcategorizes search terms and not just apple vs. Apple, but also for a more singlular term, like tricycles. After enough images, it understands how tricycles are mostly intended for kids while also making room for images of adult tricycles and motorized tricycles, according to Futurity.

Gupta believes that the sheer volume of information is the key to computer processing.

"What we have learned in the last 5 to 10 years of computer vision research is that the more data you have, the better computer vision becomes," Gupta said.

Other projects have had the same vision as NEIL -- "to create the world's largest visual structured knowledge base" -- but those required human assistance, according to Futurity. Given the 20 billion images on Facebook alone, such a project would only seem feasible with self-powered software like NEIL.

NEIL does require some human help as the software can make errors that grow exponentially. Software operators also control what search terms they want NEIL to analyze, according to Futurity.

"People don't always know how or what to teach computers," Abhinav Shrivastava, a PhD student in robotics at Carnegie Mellon. "But humans are good at telling computers when they are wrong."