The genius of Watson

“She’s afraid that if she leaves, she’ll become the life of the party.”
—Groucho Marx

What does this phrase mean? Is it funny? ? Don’t you have to be at the party in order to be the life of it? Why would someone be afraid to be the life of a party? Is she shy? Are parties living beings? Of course, being the life of the party means that you are the center of attention. So, if you are the center of attention when not present, that means that people are talking about you. From our lived experiences, we know that gossiping – people talking about you when you are not there – is generally a bad thing. Just like Sherlock Holmes, we can deduce meaning (and humor) by making these connections. But can a computer?

Watson, a supercomputer created by IBM’s DeepQA Team, is competing against two humans on Jeopardy this week. The challenge is that Watson has to interpret natural language phrases in order to determine the question for Trebek’s answer. According to David Ferucci, the principal investigator for IBM’s DeepQA team, the goal is not to replicate human thinking, but to be able to understand, process, and interact with natural language.

The genius of Watson is in making these connections….and making them quickly. Watson represents the application of decades of research into developing computers that can respond to natural language questions. It uses hundreds of thousands of processors running hundreds of algorithms simultaneously to query databases full of trillions of pieces of information. These algorithms create hundreds of possible answers. Additional algorithms search databases to find supporting evidence for the possible answers. Yet more algorithms process the answers and evidence to determine a statistical confidence that an answer is correct. Once the confidence for an answer surpasses a pre-set threshold, Watson triggers the buzzer and provides that answer. This all happens in about 3 seconds. In addition to this, Watson learns – algorithms that produce correct responses are given preference in future queries.

Watson can sort through these databases, including full original texts of Shakespeare and other classics, encyclopedias, news archives, etc. much faster than a human, but it does have one big disadvantage. Watson does not have lived experiences to include with other types of data. Often, Watson gets a correct answer, but can get wrong answers too. For example, Watson gave the answer “leg” to a question about the anatomical oddity of a U.S. gymnast in the 1904 Olympics. The correct answer is “missing leg,” but Ferucci explains, “The computer wouldn’t know that a missing leg is odder than anything else.” Ferucci continues by saying that Watson will learn this over time through additional “reading” and game playing – by learning from lived experiences.

This entry was posted in Science 2.0 and tagged , , , , . Bookmark the permalink. Trackbacks are closed, but you can post a comment.

Post a Comment

Your email is never published nor shared. Required fields are marked *

*
*

* Copy This Password *

* Type Or Paste Password Here *

You may use these HTML tags and attributes <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Subscribe without commenting