Learning By Reading And Viewing
There’s a system named NELL at Carnegie Mellon University that is attempting to learn in the way that we do when we are young: by looking at things. Since 2010, NELL, short for Never Ending Language Learning, “has been running continuously, attempting to perform two tasks each day,” according to the CMU web site. The first task is to do what many of us do every day, but on a much larger (and faster) scale: go online and read what’s on the web, by visiting millions of web pages. The second task is to improve its competency level, thus becoming a better reader, and hopefully learner, the next day.
The AP reports that in July of this year, another project named NEIL, or Never Ending Image Learning, has been trying a similar approach, but with images. NEIL, which is based on NELL’s programming, looks at photos online, and then according to its web site, NEIL has "Analyzed 5 million Images, Labeled 0.5 million images and Learned 3000 Common sense relationships."
Can Common Sense Be Taught?
An MIT artificial intelligence expert from, Catherine Havasi, says, “Understanding language in any form requires understanding connections among words, concepts, phrases and thoughts. Many of the problems we face today in artificial intelligence depend in some way on understanding this network of relationships, which represent the facts that each of us knows about the world and how words relate to one another.”
One purpose of NELL and NEIL is to try to make sense of what is seen and read, and then through using their own learned reasoning, try to define in words what is learned. Sometimes they don’t get it right. For example, NELL determined that, “english is the language of the country east_germany,” More frequently, however, NELL seems to get things correct: “Cod is a fish,” and expand on an idea: “Cod is a fish that can be served with the food fries (food) in a meal (or dish).”
NEIL is able to look at images and determine facts from them. For example, “Wheel is/has Round shape,” and “Television looks similar to Monitor.”
Not Quite There, Yet
Catherine Havasi told the AP, “Could a giraffe fit in your car? We’d have an answer,” even though we don't sit and calculate the giraffe's mass in order to be certain.
Neither NELL nor NEIL are quite there yet, to be able to determine that independently, but as processing power increases and research continues, we’re willing to bet that more advancements are to come.
