[Web4lib] Watson - IBM's "question-answering" machine (potential implications for libraries?)

Sun Feb 6 19:06:25 EST 2011

AI and question answering may, or may not, be advanced by Watson. That's open for interpretation.

What is new is the speed at which it can do it, and the comprehensiveness of the language understanding it displays. Watson approaches the speed of conversational question-answering, and begins to understand complex linguistic structures like puns or implicit meanings within the same amount of time that a human can process it.

This video is shot like a movie, which is a bit annoying, but it gives a bit more context at the real challenge behind the Jeopardy stunt.

http://www.youtube.com/watch?v=3G2H3DZ8rNc

-Andrew

On 2011-02-06, at 6:53 PM, Robert Balliot wrote:

> Bernie, I didn't know you were trying to change my mind. I am honored that
> you would care what I think.
> 
> IBM is marketing a product that can demonstrate IBM's superiority in
> supercomputing.    Jeopardy will sell advertising while mass marketing a
> participatory non-scientific evaluation that should raise its stock
> value.  Side effects may include references to John Henry and the
> Steam Powered <http://en.wikipedia.org/wiki/John_Henry_(folklore)> Hammer and
> references to products approved by the FDA that may cause drowsiness.
> 
> But, in a fair fight, I would see Jeopardy champions armed with search
> engines because that is really what is being tested - access to relational
> databases.  My aunt was on Jeopardy when Gene Rayburn hosted the show in the
> late 60s.  If she had Google's search engine available to her, she would
> have won against all comers. OMG moment -  I have watched Jeopardy for over
> 40 years!
> 
> R. Balliot
> http://oceanstatelibrarian.com
> 
> 
> 
> On Sun, Feb 6, 2011 at 6:27 PM, B.G. Sloan <bgsloan2 at yahoo.com> wrote:
> 
>>  I'm obviously not going to change Robert Balliot's mind about this, so I
>> won't continue to try. Let's just say I think "Watson" is more
>> ground-breaking than Robert thinks...
>> 
>> Interesting NY Times Magazine article about this from last summer:
>> 
>> http://nyti.ms/idJlpe
>> 
>> I know I'll be watching Jeopardy a week from tomorrow. :-)
>> 
>> Bernie Sloan
>> 
>> --- On *Sun, 2/6/11, Robert Balliot <rballiot at gmail.com>* wrote:
>> 
>> 
>> From: Robert Balliot <rballiot at gmail.com>
>> Subject: Re: [Web4lib] Watson - IBM's "question-answering" machine
>> (potential implications for libraries?)
>> To: "B.G. Sloan" <bgsloan2 at yahoo.com>
>> Cc: web4lib at webjunction.org
>> Date: Sunday, February 6, 2011, 2:28 PM
>> 
>> 
>> The New York Times used my guide to train their editorial staff on how to
>> search the web way back in 1997. My simplistic representation of boolean
>> search strategies aligned with their need for training.
>> 
>> The holy grail of AI that time *would* have been - We want a computer that
>> can understand a question based on a broad range of archived human knowledge
>> and provide the correct answer.
>> 
>> But, isn't it really  'ground breaking'  based on old paradigms?
>> 
>> Does that even seem ground breaking now given the information available
>> from structured queries in Google?
>> 
>> 
>> R. Balliot
>> http://oceanstatelibrarian.com
>> 
>> On Sun, Feb 6, 2011 at 1:31 PM, B.G. Sloan <bgsloan2 at yahoo.com<http://us.mc571.mail.yahoo.com/mc/compose?to=bgsloan2@yahoo.com>
>>> wrote:
>> 
>> Robert Balliot said: "I don't think that matching answers to questions is
>> particularly ground breaking from an AI standpoint - coming up with the
>> questions would be".
>> 
>> Well, technically, it IS coming up with questions...this is Jeopardy after
>> all. :-)
>> 
>> Seriously, though, the Op Ed piece I cited calls this project
>> ground-breaking: "Open-domain question answering has long been one of the
>> great holy grails of artificial intelligence."
>> 
>> A NY Times article from June also noted:
>> 
>> "Technologists have long regarded this sort of artificial intelligence as a
>> holy grail, because it would allow machines to converse more naturally with
>> people, letting us ask questions instead of typing keywords. Software firms
>> and university scientists have produced question-answering systems for
>> years, but these have mostly been limited to simply phrased questions.
>> Nobody ever tackled 'Jeopardy!' because experts assumed that even for the
>> latest artificial intelligence, the game was simply too hard: the clues are
>> too puzzling and allusive, and the breadth of trivia is too wide."
>> 
>> Bernie Sloan
>> 
>> --- On Sun, 2/6/11, Robert Balliot <rballiot at gmail.com<http://us.mc571.mail.yahoo.com/mc/compose?to=rballiot@gmail.com>>
>> wrote:
>> 
>> 
>> From: Robert Balliot <rballiot at gmail.com<http://us.mc571.mail.yahoo.com/mc/compose?to=rballiot@gmail.com>
>>> 
>> Subject: Re: [Web4lib] Watson - IBM's "question-answering" machine
>> (potential implications for libraries?)
>> To: "B.G. Sloan" <bgsloan2 at yahoo.com<http://us.mc571.mail.yahoo.com/mc/compose?to=bgsloan2@yahoo.com>
>>> 
>> Cc: web4lib at webjunction.org<http://us.mc571.mail.yahoo.com/mc/compose?to=web4lib@webjunction.org>
>> Date: Sunday, February 6, 2011, 12:59 PM
>> 
>> 
>> 
>> 
>> I don't think that matching answers to questions is particularly ground
>> breaking from an AI standpoint - coming up with the questions would be.
>> 
>> http://www.youtube.com/watch?v=aboZctrHfK8
>> 
>> R. Balliot
>> http://oceanstatelibrarian.com/contact.htm
>> 
>> 
>> 
>> On Sun, Feb 6, 2011 at 12:26 PM, B.G. Sloan <bgsloan2 at yahoo.com<http://us.mc571.mail.yahoo.com/mc/compose?to=bgsloan2@yahoo.com>>
>> wrote:
>> 
>> From the New York Times:
>> 
>> "I.B.M.’s groundbreaking question-answering system, running on roughly
>> 2,500 parallel processor cores, each able to perform up to 33 billion
>> operations a second, is playing a pair of 'Jeopardy!' matches against the
>> show’s top two living players, to be aired on Feb. 14, 15 and 16."
>> 
>> I'm definitely going to be tuning in to these Jeopardy episodes. I'm
>> curious to see how well "Watson" does against two smart humans.
>> 
>> Can't help but wonder about long-term implications for libraries, say, ten
>> years down the road? What if we had sophisticated affordable
>> "question-answering" machines in ten years? What would that mean for
>> libraries?
>> 
>> For some background, here's a link to an Op Ed piece in today's New York
>> Times: http://nyti.ms/hXaoWX
>> 
>> Bernie Sloan
>> 
>> 
>> 
>> 
>> 
>> _______________________________________________
>> Web4lib mailing list
>> Web4lib at webjunction.org<http://us.mc571.mail.yahoo.com/mc/compose?to=Web4lib@webjunction.org>
>> http://lists.webjunction.org/web4lib/
>> 
>> 
>> 
>> 
>> _______________________________________________
>> Web4lib mailing list
>> Web4lib at webjunction.org<http://us.mc571.mail.yahoo.com/mc/compose?to=Web4lib@webjunction.org>
>> http://lists.webjunction.org/web4lib/
>> 
>> 
>> 
>> 
> _______________________________________________
> Web4lib mailing list
> Web4lib at webjunction.org
> http://lists.webjunction.org/web4lib/
> 