New Computer System Uses the Internet to 'Learn'

Dennis Faas's picture

Computer Science researchers at Carnegie Mellon University have created a computer system capable of performing tasks never thought possible. Dubbed "NELL" (short for Never-Ending Language Learning), the computer system has the ability to learn, read and express assumptions on a variety of topics.

Researchers were provided funding from Google and DARPA agency (a research and development arm of the U.S. Department of Defense) to create a system capable of understanding the entirety of the human language. The Carnegie Mellon group programmed NELL to categorize the information it finds on the Internet and create links between them, much the same way as the human brain is able to do.

Computer Learns Through Inferences

The manner in which NELL is able to learn through inferences is remarkable, as depicted in the following example:

NELL understands that Peyton Manning is a football player (one category). Nell understands that the Indianapolis Colts are a football team (another category). In scanning the textual patterns across the Internet, NELL can infer (with high probability) that Peyton Manning plays for the Indianapolis Colts -- even if NELL has never read the actual sentence "Peyton Manning plays for the Indianapolis Colts". (Source: switched.com)

NELL Not Yet Perfect

As you can imagine, NELL makes some mistakes in its sorting criteria, an issue that can actually lead to some surprisingly hilarious results.

According to NELL, "Internet cookies" are classified as "baked goods", a "security risk" is classified as another version of the popular board game "Risk", "The Right to Bear Arms" is classified as a hobby and "Klingon" is classified as an ethnic minority (which may not be that much of a stretch).

As it stands, NELL is about 74 per cent accurate overall in its assumptions, which is much more precise than anyone could have ever expected from the emergent technology. (Source: canoe.ca)

The Carnegie Mellon group has also expressed their desire to create an autonomous system, but as it stands, NELL is "not there yet".

Rate this article: 
No votes yet