Hitting the Books

Reading comprehension is a skill that measures how a person is able to grasp and understand what they are reading. Improving this involves reading and lots of practice to gauge and discern differing usage and contexts of words. The same thing applies for artificial intelligence.

Facebook released several data sets that it has been using to train its home-grown neural networks. Some of these are texts of classic novels, such as Rudyard Kipling's The Jungle Book, J. M. Barrie's Peter Pan, and Lewis Carroll's Alice's Adventures in Wonderland; others are collections of fairy stories, such as Andrew Lang's Fairy Books, Nathaniel Hawthorne's Twice-Told Tales, and Oscar Wilde's The Happy Prince and Other Tales. Most of these were obtained from the free online library Project Gutenberg.

By training its AI on these books, Facebook is trying to ensure that its AI is able to process what it's reading and is able to make logical connections. Central to this is what Facebook calls the "Children’s Book Test," which it uses as a metric to ascertain the progress of the AI.

In their paper, Facebook researchers trained a neural network using a sample of books from the list. After this, they presented it with short excerpts from the stories it had not read. The AI was then asked to choose a word from a list of options which would fill the gap left by a missing word in the final sentence.

Reading Between the Lines
The AI chose “Gryphon” to fill in the gap. Credit: Facebook

By being able to answer such questions, the researchers believe that it demonstrates that an AI can make decisions or judgements by drawing on a situation’s wider context. This skill is crucial in representing and remembering complex pieces of information, and it would improve the AI's performance in situations requiring such processing outside of reading; another Facebook-devised intelligence test is geared to comprehending the relationships between ideas in short stories.

“Our team taught the computer to look at the context of a sentence and much more accurately predict those more difficult words—nouns and names—which are often the most important parts of sentences,” Facebook’s CEO Mark Zuckerberg explained in a post.

“The computer’s predictions were most accurate when it looked at just the right amount of context around relevant words—not too much and not too little,” he said, calling it the “Goldilocks Principle.”

These results are being incorporated into M—the digital assistant currently being tested within Facebook Messenger—although it will take more research and processing before the AI is able to understand and respond to open-natured inquiries as well as humans do.
But what matters is that computer programs are gradually learning those cognitive behaviors that come so naturally to human beings; it's a difficult process (perhaps more difficult than anyone imagined) but the programs are steadily evolving. The ultimate goal is artificially-intelligent (or approximately so) programs and algorithms that can interact more naturally with humans—think of Siri or Cortana, but with the ability to understand language inflection, or such subtle flavors of thought as irony, sarcasm, impatience, etc.
And as we learn how to incorporate intuition and cognitive sophistication in our computer programs, maybe we'll learn a little more about ourselves along the way.

Share This Article