addressalign-toparrow-leftarrow-rightbackbellblockcalendarcameraccwcheckchevron-downchevron-leftchevron-rightchevron-small-downchevron-small-leftchevron-small-rightchevron-small-upchevron-upcircle-with-checkcircle-with-crosscircle-with-pluscrossdots-three-verticaleditemptyheartexporteye-with-lineeyefacebookfolderfullheartglobegmailgooglegroupsimageimagesinstagramlinklocation-pinm-swarmSearchmailmessagesminusmoremuplabelShape 3 + Rectangle 1outlookpersonJoin Group on CardStartprice-ribbonImported LayersImported LayersImported Layersshieldstartickettrashtriangle-downtriangle-uptwitteruseryahoo

The Atlanta Science Tavern Message Board The Atlanta Science Tavern Discussion Forum › Watson versus the Jeopardy! champions

Watson versus the Jeopardy! champions

This message board is read-only.

Ernie C.
user 7596470
Atlanta, GA
Post #: 22
I suppose many people have now seen or at least heard about IBM's Watson computer and its face-off with two human Jeopardy! champions. Recently, I saw a very nice Nova episode about the technology involved, which you can also see by going here:­

My thoughts on this, after watching the Nova episode, are:

* While Watson is impressive technology (on the hardware and software side both), it is not really all *that* impressive. The Nova program doesn't delve into details about the methods that Watson uses, but it looks to me as though it doesn't try to do any deep contextual analysis of the Jeopardy! clues; rather, it just focuses on a few key words and looks up the key pieces of evidence to support potential answers (ahem... potential questions).

* I would have thought that the kind of wordplay in Jeopardy! clues would have been too difficult for a computer to handle. Apparently, zeroing in on the few key details in a Jeopardy! clue (or crossword puzzle clue for that matter) is something amenable to simple brute-force statistical techniques after all.

* We have had a computer beat the world chess champion, a computer hold its own against two Jeopardy! champions, and cars that can drive themselves in cities (the DARPA urban challenge). What is the next milestone for AI, and how will it be crossed? It seems to me that each new milestone requires not only more processing power to exceed, but also the integration of several different algorithms to solve the various subtasks involved. It is as though we are moving incrementally towards machines that have what Marvin Minsky called ``Society of Mind'', referring to the fact that a human mind is a conglomerate of a vast assortment of submodules operating interdependently to handle tasks like vision, language processing, thinking, etc. (I do think, however, that just combining together methods we already know about will not be enough to achieve something like a functioning general-purpose AI -- theory will be important as well.)
Ernie C.
user 7596470
Atlanta, GA
Post #: 23
Perhaps Watson is more sophisticated than I gave it credit. The wiki article has some discussion of some of the technology involved:


It mentions that ``thousands of algorithms'' were employed, everything from advanced natural language processing (whatever `advanced' means), to knowledge representation and reasoning algorithms.

Jeffrey Shallit, who maintains the blog Recursivity, had some very positive things to say about the technology (and some negative things to say about its critics):


I wonder how soon until spin-offs of the technology hit the medical profession, one of IBM and Nuance Communications' stated ``next steps''...
Lawrenceville, GA
Post #: 95
Powered by mvnForum

Our Sponsors

  • Abrupt Media

    Abrupt Media helps businesses increase visibility on the internet

People in this
Meetup are also in:

Sign up

Meetup members, Log in

By clicking "Sign up" or "Sign up using Facebook", you confirm that you accept our Terms of Service & Privacy Policy