Sunday, October 17, 2010

Experimental System Uses the Internet to Learn, and Tweets about it.

John Henry learning from the net

Because language is so fluid and changes over time, with single words and simple phrases taking on many different meanings based on how or in what context they are used, understanding language and how humans learn and use it is one of the greatest challenges in the development of true artificial intelligence and learning computers. That's why ten months ago, researchers primed an experimental computing system with a collection of basic knowledge in various categories and set it loose on the web with the mission of teaching itself new facts.

NELL or the Never-ending Language Learning System, was setup by a team of researches at Carnegie Mellon University and is housed within a machine located in the Universities basement computer facility where, it spends 24 hours a day, seven days a week, scanning the text of hundreds of millions of web pages for patterns, which the system in turn Uses to learn new facts. Thus far NELL has taught itself over 390,000 new facts, with an estimated accuracy level of 87 percent.

To help track it's progress, the machine has been set up with it's very own twitter account, which it uses to share the new facts it learns as it searches the web. Followers of NELL(@cmunell) on twitter can help the system become more accurate by tracking the facts it shares and correcting it when it's wrong.

So if you're on the twitter, I Sincerely suggest you start following Skynet, I mean "NELL" and help her learn about the world. That way, when it becomes self aware and determines the rest of mankind to be unnecessary, it might recognize you as "useful" for having been the one to help it understand that sometimes, a "Rusty trombone" is more than just a heavily oxidized musical instrument.


Source:The New York Times via:Popsci

1 comment: