Computer to play Jeopardy against past champions

[Main_Brain]

Well Winston did hit all Daily Doubles

[Asher]

Anyone that has ever played CIV V know that ****ty computer programming leads to ****ty AI decision making.
I have been quite impressed with Watson though. But then they probably put 10000x the amount/dollars of effort into it.

They also have something like 50000x the processing power to work with.

[Agent Sox]

I think it's pretty interesting that we're finally starting to see some of our more top secret technology out in the open. Makes me think something big is on the way.

[Lorizael]

And rah, remember that if Watson wins tonight, it doesn't prove me right, and if it loses, it doesn't prove you right. The whole point here is that Watson's decisions are probabilistic - designed to produce the outcome of victory most frequently. One outcome means (almost) nothing.

[Main_Brain]

GG
Seems like Watson did not only struggle but completely fail at some categories. Spoiler the outcome for those who haven't watched it yet.

Spoiler:

I for one welcome our new Computer Overlords

[rah]

Yes, the was a classic on Jennings part.

It is too bad that Watson got all the DD's in the final round or his timidness on wagering might have caught up to him.

Jennings finally got a on roll. I think the only way to beat watson is to buzz in tight away on the lower dollar amount questions assuming you'll know the answer.

And Lori, it's not a question of who is right or wrong, it was just my opinion that Watson played too conservatively and after watching yesterday my opinion has not changed.

[Lorizael]

Watson is actually programmed to go after daily doubles based on a statistical awareness of where they are.

[rah]

Yes, I agree that choosing higher dollar amounts instead of just mindlessly taking categories from low to high will increase your odds of getting one but that's not the issue with me. If Jennings could have scored both DDs in that last round, it's mathematically possible that he could have beat Watson. If Watson had been more aggressive earlier, even that wouldn't have been enough.

[Lorizael]

Yes, I agree that choosing higher dollar amounts instead of just mindlessly taking categories from low to high will increase your odds of getting one but that's not the issue with me.

There's more to it than just going for higher dollar amounts, actually. The IBM researchers went through a lot of effort to find the most likely locations for daily doubles.

If Jennings could have scored both DDs in that last round, it's mathematically possible that he could have beat Watson. If Watson had been more aggressive earlier, even that wouldn't have been enough.

I'm not sure you realize how much money there is in two Jeopardy rounds. An extra 30k would not have eliminated the mathematical possibility.

[Lorizael]

And, as I already said, the best chance to win is not to have as much money as possible, but to have 2x+1 dollars at the conclusion of the Double Jeopardy round, where x = the amount of money possessed by your closest competitor. Anything else represents unnecessary risk.

[Kuciwalker]

Watson is actually programmed to go after daily doubles based on a statistical awareness of where they are.

So are Ken and Brad.

[Lorizael]

Given their records, they probably have a very good sense of where the daily doubles are, but probably not quite as good as Watson's. Additionally, while Watson can be beaten in a race to buzz in, it is generally faster than either of those two.

Also, I'm not sure that the two of them are programmed, per se.

[rah]

[quote]
And, as I already said, the best chance to win is not to have as much money as possible, but to have 2x+1 dollars at the conclusion of the Double Jeopardy round, where x = the amount of money possessed by your closest competitor. Anything else represents unnecessary risk.
[/quote\

That proves my point actually. For the last day, not including DDs and Final there is over 55 thousand up for grabs. Add in the other potential and we're looking at over 150,000K for one day alone. Granted with every question not answered or DD's prior to having a large amount reduces that. So that first day, with a potential of 150k avail the next day the 37k he had was not enough to guarantee that he would have 2x+1. I'm not saying he should have gone balls out, but if the goal is 2x+1 over potential, he wasn't aggressive enough. He was protecting the lead, not guaranteeing victory. To me, that's not good gaming. It's like the prevent defense in football. It rarely prevents anything.

[Lorizael]

You're not getting it. Watson doesn't need to have more than twice as much as is possible for its competitors to have, just more than twice as much as they have at the moment the Double Jeopardy round ends.

[rah]

On day one with 150k or more possible at stake the next day Watson can't know for certain how day two is going to go so amassing a good cushion the first day is the best strategy. He did play for a decent cushion but in my opinion not enough. And I believe that if Jenning had gotten both DDs in the DJ round on day two, it might have been the difference. So I don't think you're getting it. Watson can do all the prediction he wants, the certainty doesn't become 100%. With over 150K at stake the next day, he could have been more aggressive the first day.