Victor Shulist - Feb 15, 2011:

I can’t remember what Ken’s score was, do you remember?

Yes, it was $2,000.

Awesome thanks.

Thunder Walk - Feb 15, 2011:

Victor Shulist - Feb 15, 2011:

I liked Watson getting a daily double pretty much right off the bat. He’s smart *and* lucky!

That was the biggest surprise of all. Watson always followed the pattern of picking questions in each category in the traditional manner from the top (easiest questions) down, with that one exception. The Daily Double turned out to be a mid-level question WATSON picked by skipping the ones above it.

Was hitting it so early in the contest a statistical guess at where it might be based on past games, and part of the programming?

It also made me wonder if part of the programming contains additional elements that do more than make an effort to be better than humans at answering questions. Is there a component to the programming that strategizes on winning a greater amount of money over answering the most questions correctly? Is it, in the end, not only an effort to overcome language, but merely a chess match?

Yes, there are Jeopardy strategies. You can read about some of them here. I think I recall from one of the interviews that Watson is using some of them.

@Victor - Did you notice that Watson is not using grammar?

Merlin - Feb 15, 2011:

@Victor - Did you notice that Watson is not using grammar?

Most of the video footage I’ve seen indicates grammar plays a role.

http://www.theepochtimes.com/n2/content/view/51179/

“Watson is designed to react to aspects of the English language, such as grammar, pop culture references, colloquialisms, and even idiomatic phrases”

So of course Watson has to deal with far more than grammar.

There seems to be a belief on this site that my approach is grammar and only grammar.  The truth is, my philosophy is to give my bot grammar rules, and not bother with having it learn them (it’s cool if it were to learn them, but I see that as not absolutely necessary).

Victor Shulist - Feb 15, 2011:

http://www.theepochtimes.com/n2/content/view/51179/

“Watson is designed to react to aspects of the English language, such as grammar, pop culture references, colloquialisms, and even idiomatic phrases”

So of course Watson has to deal with far more than grammar.

There seems to be a belief on this site that my approach is grammar and only grammar.  The truth is, my philosophy is to give my bot grammar rules, and not bother with having it learn them (it’s cool if it were to learn them, but I see that as not absolutely necessary).

Victor,
I was not referring to your bot. I was referring to the difference in grammar centric versus concept or phrase centric approaches.  In some ways, any bot that looks for a pattern of more than one word will be dealing with grammar.

In what I understand of the Watson approach, it does not try to parse and understand the complete sentences of the questions. Instead it takes sentence fragments (grammar fragments?) and attempts to build a list of potential answers. Combining answers gives it a relative score, and if that score reaches a threshold it buzzes in.

In what I understand of the Watson approach, it does not try to parse and understand the complete sentences of the questions. Instead it takes sentence fragments (grammar fragments?) and attempts to build a list of potential answers. Combining answers gives it a relative score, and if that score reaches a threshold it buzzes in.

That’s also how I thought it worked.

It appeared that WATSON was cleaning up last night, but this morning, various reports indicate the other players were actually, “locking themselves out.”

“... contestants have to wait until Alex Trebek is completely finished reading the question before they are allowed to buzz in. Buzz too soon, and your buzzer is “locked out” for a quarter of a second, giving opponents the chance to jump in and answer before you. Contestants who wish to buzz in as fast as possible must either try to guess when the buzzers will activate (risking getting locked out if they are too early) or rely on their reflexes to buzz in when they see the lights (risking having reflexes that are too slow and allowing another player to buzz in first.”

According to Trebek on one of the morning TV programs, it occurred nine times last night.

Erwin,

May I ask…
Have you heard back from IBM yet ?

Tell IBM that Watson inspired me to build
a collective, or an integration of several
open source A.I. engines:

1. Eliza
2. Program E
3. Program O
4. MegaHAL
5. ... Other A.I. engines to be assimilated.

You can mention: http://79b.org to IBM.

keywords:  7 of 9, The Borg, Star Trek Voyager

Thank you, Erwin.

haven’t contacted them, I’m sorry, it has been so busy lately. So many invisible things going on. This will change soon, very soon.

Looks nice 8pla. At first, though, I thought, it was giving an error (with the array), but then I read it it.

Yup, very neat idea, 8pla. Do you have any plans to sort or rank the “collective” answers, as the number of bots grows? Perhaps users could be asked to indicate which answer they like best? The bot could then take advantage of the strengths of each member of the “collective”.

Why did Watson bet odd amounts?

[ul][li]http://www.wired.com/geekdad/2011/02/whats-with-watsons-weird-bets-and-other-questions-about-ibms-jeopardy-playing-machine/[/li][/ul]

Watson bet $1246 on one Daily Double and $6,435 on the other — Lally said the Watson team did include game theorists, who programmed winning strategies (like searching the board’s higher-priced clues early for hidden Daily Doubles) into Watson’s game plan. But though human players usually pick round numbers when naming their Daily Double wager, there’s no reason why they have to — so the team decided to let Watson choose its own amount based on its own algorithms.

As for why humans were inadvertently locking themselves out…

Although on Tuesday night it looked like Watson had an advantage over his human opponents when it comes to “buzzing in,” Welty said that wasn’t the case. Watson is programmed to wait until the light next to host Alex Trebek is lit, alerting players that they are allowed to press the buttons that signal their desire to respond (buzzing in early gets a player locked out momentarily). Human players get a split second head start, he said, because they are listening for Trebek to finish reading the clue rather than waiting for the light.

But, I don’t understand why the humans wagered such a small amount for the Final Jeopardy question, rather than their entire winnings, since it wasn’t really a game that would continue, and the only goal for the humans was to beat Watson.  They had nothing to lose.  Watson’s main goal was to prove it could process natural language.

Found a great article talking about Watson’s computing power and IBM’s hope to use Watson in hospitals. From the article:

The original compute algorithm single threaded on a single core processor took two hours to scan memory and produce an answer to a question.

Gives me hope that my own “small scale” NLP efforts are still tenable as a proof of principle. After that, it’s apparently all about throwing CPU’s at the problem.