Monday, February 14, 2011

IT BECOMES SELF-AWARE AT 2:14 AM EASTERN TIME AUGUST 29TH: Some spoiler-free thoughts on tonight's first installment of the Jeopardy: Man v. Machine challenge (though spoilage is fine in the comments):
  • Obviously, not every Jeopardy! viewer has been following the media coverage of Watson and its background, but did we really need so much background/documentary on Watson in the episode? We came to watch the game--not a briefer version of the Nova episode.
  • The "show Watson's calculations" format is amusing in that it shows us Watson's sometimes bizarre answers, but it takes away from the "play along" aspects of the game.
  • The "2 games over 3 days" format makes me wonder how much we're getting a narrative edit in the episodes.
  • One thing Watson clearly cannot do is listen to his competitors' answers and draw conclusions/guesses from them to put together the best possible choices/answers. Had there been even a rudimentary function to do that, there were at least a couple of clues Watson would have gotten right and didn't.
Discuss below, and we'll be back with more coverage over the next few days.

13 comments:

  1. GoldnI8:31 PM

    The biggest thing I took away from it is that it's really only a matter of time before the computers become sentient and take over.

    ReplyDelete
  2. Matt B8:48 PM

    30 minutes for one round of Jeopardy was definitely a disappointment.

    ReplyDelete
  3. I agree that there's too much background: give us a special on how they made Watson, then give us three regulation games of Jeopardy.

    On a side note, Watson's affinity for switching categories is ruining my ability to play at home...I can't keep track of which questions belong in which categories.

    ReplyDelete
  4. mhgatti9:14 PM

    All that background, and I don't remember seeing anything about how they were compensating for Watson not having to "buzz in." Also, they didn't mention how "he" was deciding what categories to choose or any kind of wagering strategy (which came up quickly with an early daily double).

    ReplyDelete
  5. I noticed that the early daily double- Watson's first selection- was in the spot the Slate article identified as the most likely location for the daily double, then Watson went down from the top. Possibly trying to avoid others from building a lead?

    ReplyDelete
  6. Well, wagering "strategy" doesn't really come into play in that situation.  My standard play method when playing on a computer or console is to start in the middle of the category to gauge difficulty.

    ReplyDelete
  7. Watson's not being fed others' wrong answers was a surprise indeed.  Nor did it seem like the humans figured out what Chuck did here a few days ago: stick with categories more prone to wordplay, and away from straight trivia.

    ReplyDelete
  8. It wasn't just Watson that didn't seem willing to play with the wrong answer information.  I was shocked Rutter didn't put together the prior wrong answers on the gymnast question and come up with "had only one leg" as an answer, since all the pieces were there.

    ReplyDelete
  9. Tosy and Cosh12:23 AM

    They did indicate that Watcon was pushing a physical buzzer, just like the contestants - the computer is attached to a device that can mechanically depress the button on a standard buzzer. Not sure if they compensate reaction times or not (I'm assuming a computer can read faster than any human, no?). I was curious more about fairness/logistical stuff like this than the AI background, really. Does Watson have access to decades of Jeopardy questions to cross-reference?

    ReplyDelete
  10. Jenn C5:26 AM

    I enjoyed Trebek's snide "no...Ken already said that" response when Watson buzzed in after Ken already gave the wrong answer.

    ReplyDelete
  11. mhgatti11:05 AM

    <p><span>More on answer selection:</span>
    </p><p> 
    </p><p><span>If Watson gets to choose a category and clue, its first priority is finding any remaining of the three Daily Doubles in a game. These clues allow a contestant to wager a specific dollar amount on the clue without worry of the other two contestants buzzing in. Jennings, Rutter and Watson have a high chance to answer these correctly, so Daily Doubles provide three opportunities for a critical score boost.

    The Watson Research team studied the historical distribution of Daily Doubles and found they appear most-frequently in the three bottom rows, with the fourth being the most common. Daily Doubles also most frequently appear in the first column. Watson also makes use of even more statistics to dynamically predict their location based on what has been exposed so far in a game.</span>
    </p><p><span>Once the Daily Doubles are off the board, Watson looks for the lowest clue value in a category, for which there are still a significant number of high value clues. Lower value clues help it get the gist of a category with less risk, so that it has a better shot at the high value clues to come.</span>
    </p><p> 
    </p><p>http://ibmresearchnews.blogspot.com/2011/02/knowing-what-it-knows-selected-nuances.html</p>

    ReplyDelete
  12. That's interesting that Watson bases answers in a category off of previous question-answer pairs in a category.

    It was also interesting that Watson was outbuzzed by Ken a few times.

    Watson definitely did far worse in wordplay than trivia. The wordplay performance didn't seem that impressive. I think Colby Cosh has the right take: http://www2.macleans.ca/2011/01/16/ill-take-cheap-publicity-stunts-for-1000-alex/

    ReplyDelete