Watson & the Future

By Jason Poggioli

At times it can be difficult to remember just how blazingly fast technological progress is moving. But then there are times it just leaps up and screams, “You’re living in the future right now and you’d better pay attention!” That’s how I felt this week watching a computer compete on Jeopardy against arguably the two best Jeopardy contestants ever, Ken Jennings and Brad Rutter.

I’m not sure it’s possible to overstate the importance of this milestone. A machine has been created that is capable of parsing Jeopardy clues and giving correct answers. Clues that often have double meanings which cannot be easily understood by a computer. No special accommodations were made other than that clues were fed the machine via a text file rather than verbally. The machine, named Watson by its IBM creators, even had to mechanically press the button, just like Jennings and Rutter.

Successfully deciphering the clues is IBM’s giant accomplishment because it’s not like you can take a Jeopardy clue, type it into the Google search engine, and get an answer. Although the power of Google may seem at times like the ancient Greek Oracle, it’s only an indexed database of raw information. Typing in the actual Jeopardy clue would yield nonsensical results because it would simply look up web pages that contain the words you typed. Taking the entire clue and correctly surmising its meaning has been the stuff of science fiction until now. Besides, Watson wasn’t even connected to the Internet for this competition.

When faced with a question, Watson pulls apart the words to understand how they relate to one another within the clue. Ordinarily a computer must be fed a strict list of instructions – a program – written by people. Watson’s great leap forward is that, although it still requires instructions, the resulting program allows it to be fed naturally worded clues that can be understood well enough for it to find correct answers in its database. Put another way: Instead of people needing to learn the computer’s language, the computer has been taught to understand ours. The entire process is called “natural language understanding,” a specific field of computer science in which programs are written to successfully understand human language. It also happens to be the field of computer science the company I work for specializes in.

The practical application for Watson is its ability to digest large quantities of information and find relationships or patterns buried within the data that – due either to its complexity or immensity – escapes human notice. For example, patterns of disease cross referenced with geographic location could unearth an unknown toxic environment. Similarly, astronomical data could be fed into it and out would pop a newly discovered stellar phenomenon. Maybe the underlying cause of the cicadas’ 13-year swarm patterns could be found.

To be fair, it’s probable that Watson’s program has been specifically written for understanding the short succinct clue styles found on Jeopardy rather than entire sentences. So it’s not as if we have to worry about computers taking over the world – yet. Additionally, Watson doesn’t have a context for the clue – it doesn’t understand the clue like you or I would. It may be able to sweep the category of “Characters found in Beatles songs.” but the clues it reads aren’t going to get it humming “Maxwell’s Silver Hammer.”

Although… When a computer becomes so good at understanding human language that you can’t tell if it’s a computer or a person, isn’t that a clear definition of artificial intelligence?

Alan Turing, a computer scientist, thought so, and the definition I just described is actually a test that’s named after him. If a computer can pass what is known as the Turing Test it’s said that you could be typing at a terminal having a conversation with either a computer or real person on the other end and not know the difference. It’s a popular competition among computer scientists to create the program that could successfully have an open ended conversation that passes as a human. With Watson that goal has become much more achievable.

The future is now.

7 Responses to “Watson & the Future”

  1. CarrieJacobson Says:

    Fabulous column on a fascinating subject. I have watched the first two days of this Jeopardy! challenge with true wonder and awe. It is compelling to watch that computer, see what it gets right, what it gets wrong, and to try to understand what IT is understanding. Wow. So glad you wrote this!

  2. Jo Galante Cicale Says:

    one of the best articles! my husband and I didn’t quite understand how Watson was operating. thanks for shedding light.

  3. Guy Hathaway Says:

    Proposal for Phase 2:

    Require all would-be candidates for positions of authority to demonstrate that they can parse and respond to the same set of questions of fact as clearly and thoroughly as Watson can.

    If they can’t, would we want them to “represent” us?

    (Is there a flashing “BS” light on Watson?)

  4. Guy Hathaway Says:

    Waiting.

    Waiting.

    Waiting.

    Thanks, anyway, for the intriguing article.

    Goodbye

  5. Jason Says:

    Jo, glad you liked the article!

    Guy, a “BS” indicator actually isn’t all that far fetched. Web sites like FactCheck.org already have humans pouring over political speeches and rhetoric attempting to distill the truth from the hyperbole. With something like Watson and a loaded database of actual political facts that entire BS detection process could be automated.

  6. Geoff Howard Says:

    Great article but how do we know who really wrote it, you or Watson?

    Seriously, that’s really interesting and I hover between excitement and fear. Let’s say conservatively that they – whoever they are – make as many advances in the next 10 years as they have in the last 10, what will be the possibilities? Will it – in 2032, say – be able to take every facet of my life, all my accomplishments and failures (stored in its database) and craft and deliver a great eulogy at my memorial service, something absolutely eloquent and perfectly attuned to the sensibilities of the people in the audience, sensibilities that are likewise stored in its db?

    And picking up on a previous comment by Guy: could Watson have a built-in lie detector that worked off of voice stress? Now we’re talking scary.

  7. Emily Lee Says:

    Not only will it be able to deliver a stunning, and tender, eulogy, Geoff, but at the triumphant conclusion it will whisk away a velvet drape to reveal the GeoHow2032, a magic eight ball guaranteed to respond to queries as only you could.

    great article, Jason!

Leave a Reply