There are clearly bad categories; it whiffed on every "decades" question.

I enjoyed that. The main thing I've learned so far is Ken Jennings REALLY dislikes getting owned.

I noticed Ken and Brad were able to get their timing right in the 2nd half of the episode... what changed? can't the computer win that literally every time?

i have never seen him more downcast than during that interview

i hope they give it a before and after category too

I agree that there are bad categories for watson but I don't know if the algorithm is built in to purposely avoid those categories. Like how many questions does watson have to miss or not have sufficient confidence in time in order to consider a category a "bad one".

i dunno, seemed odd that the computer ran train for the first half then just imploded

It could be that Watson actually took longer in the 2nd half trying to get its confidence level on a question past the marker due to more difficult questions.

Sure, I see what you're saying.

that was awesome, I wish they didn't have the silly vignettes though and we were getting 3 full games.

In terms of buzzer, the nyt article speaks to how the best humans are faster than the machine (at least they were months ago when the article was written: http://www.nytimes.com/2010/06/20/ma...pagewanted=all )

Although we didn't really see it get beat much on the buzzer, Ken beat it twice (Julia and leg)

One that I found really interesting was one where it didn't buzz in during the decades category (16 mins into the 20 min ep for fellow dl'ers), where none of three most likely answers were even decades.

I would hope (and would guess) that the writers didn't know which clues were going to get on the Watson eps when they were writing them. I remember during Ken's huge run, the writers said they couldn't write clues with Ken in mind even if they wanted to, since they write the clues so long in advance and don't know when a particular category will be used.

Think part of the reason its not 3 full games is that incase watson just utterly dominated or got utterly dominated, the other 2 games would not be that fun.

so wait is it just one game or two

Two. I imagine they'll have moved past the background and stuff tomorrow while doing Double/Final of game 1 and Wednesday will be one self-contained game.

I'll be pissed if this ends a 1-1 tie.

I'd been assuming it's standard Jeopardy tournament final format, where the scores add. Maybe not?

EDIT: Also, for whatever reason, I've been finding that I'm a lot more psyched about this than I was about Deep Blue back in the day. It probably helps that it's easier to follow.

here's a question: does watson know the other players' scores? i guess it must, but that might be why it is fishing for DDs earlier, that way its betting is (likely) closer to optimal

surely it has to or else final jeopardy could be hilarious

yea it must know the scores for final and almost certainly DDs too.

A more interesting question is whether the scores impact the threshold. We know that Watson has that threshold of certainty for buzzing in. If it's down a lot with not much time left in the game, it seems that threshold should decrease and vice versa, but there's no indication of that being the case. (e.g. if there's one answer left in the game, does Watson realize when it needs that $ to even have a chance in Final?)