Computers are getting smarter at an intimidatingly fast pace. We’ve seen computers that can do things from creating video games to repairing themselves. It's only a matter of time before computers get smarter than us, but hopefully not to the point that they say, “I’m sorry, Dave. I’m afraid I can’t do that.” A new program developed by computer scientists at the University of Washington and the Allen Institute for Artificial Intelligence in Seattle can search the internet for any topic and virtually learn everything about it.
Why do we need a program like that? When you’re looking for specific topics on the web, you can spend hours going through all the data, because, well, there’s a lot of it. This new program can do all that hard work for you, in the fraction of time it takes for you to do it. Called LEVAN (Learning EVerything About Anything), this program searches through millions of Google Books and Internet images and learns everything it can about a specific topic. Once it's done, it gives you a list of results to help you understand more about the topic you chose.
What makes this program so unique is the way it searches for visual concepts. When searching through images, it doesn’t just look at the image’s caption for words related to the topic. The program’s algorithm can actually look at pixels in any image and recognize specific shapes. For words, it searches through all English-language Google Books and uses the algorithm to recognize related text. It collects text that denotes visual concepts: for example, when looking for details about horses, it would collect information about a “jumping horse” or a “barrel horse,” but not for “my horse” or “first horse.”
Right now, the program has a library of around 175 topics, including everything from “airline” to “beautiful” to “innovation” to “robot.” If the subject you’re researching isn’t in its existing database, you can have the program search for it. However, you might have to wait awhile: some concepts might keep the program busy for up to 12 hours.