The thing I didn't like about the match was the fact they had to input with a keyboard then the comp responded. I also wonder if the comp did the selection of categories. nobody mentioned that. It blew them away the first day, won bigtime, that was what allowed the win. I wonder if there will be a rematch?
After the Kasparov win they dismantled the computer, think they would do the same here?
I would like to see a match where the input comes from machine recognition of natural language which did not happen in this match.
It's too bad that in a lot of questions where the computer was slower to click the button, it would have gotten the question wrong, so in a sense the speed of the humans worked against them at least in double jeopardy.
Originally posted by sonhouse1. Yes, audible speech recognition is the next hurdle, but see comment 4 below. Typed inputs will do for the business applications, for a while.
The thing I didn't like about the match was the fact they had to input with a keyboard then the comp responded. I also wonder if the comp did the selection of categories. nobody mentioned that. It blew them away the first day, won bigtime, that was what allowed the win. I wonder if there will be a rematch?
After the Kasparov win they dismantled the compute ...[text shortened]... wrong, so in a sense the speed of the humans worked against them at least in double jeopardy.
2. I seriously doubt that Jeopardy would deviate from their in-house selection of categories. It would be a serious PR problem if it is suspected that they let contestants pick categories, especially IBM. Check out http://en.wikipedia.org/wiki/Quiz_show_scandals
3. Rematch? Only if there is a potential market advantage for comment 4. They will work on the flaws -- Toronto? -- and demonstrate that they are cured.
4. I am sure they are looking for ways to market Watson. This one was just Watson 1.0.
5. You didn't ask, but I wonder if the humans might have been able to adjust their game to beat Watson, over time, as I have been able to do against a particular on-line Chess machine.
6. Watson did a lot of high value bets right off the bat. I wonder if it will change human betting on Jeopardy.
Originally posted by JS357I'm pretty sure sonhouse meant choosing which of the six categories presented to get an answer from during the game. I don't want to put words in his mouth or speak for him, but this seems pretty obvious to me. I don't see how you could think he meant IBM selecting which categories would be in the game.
2. I seriously doubt that Jeopardy would deviate from their in-house selection of categories. It would be a serious PR problem if it is suspected that they let contestants pick categories, especially IBM. Check out http://en.wikipedia.org/wiki/Quiz_show_scandals
There is a part of all this that is just plain gimmicky. I'm referring to the dramatic sales campaign aspect. IBM is getting a HUGE amount of free publicity for this. I don't mind that, it's the US free market at work. Also, it's not "free" as IBM had to develop the computer and they heavily advertized the event. I presume that they were heavy advertizers during the shows (I didn't watch).
Here's where I'm going:
The show doesn't need to be live and concurrent.
It's a computer program! IBM can test it's computer against the humans endlessly. They can take recordings of any number of shows and enter the computer questions just as they did with the "real" show. The computer's responses would come just as fast or slow, and just as right or wrong, or silly, or stupid, or brilliant, as with the televised show.
Now, to be real. I presume that is exactly what IBM did extensively before they ever set up the show. IBM knew exactly how fast and how right or wrong or clever or stupid the computer would be. It is revealing that the computer and the very best Jeopardy players were so evenly matched! Does anyone imagine that was just an accident? The most likely answer is that IBM did the set up at exactly the point where they developed the computer to match evenly with the very best players, which led to the most exciting TV viewing.
Having said that, I do find it interesting where the state of the art of aritificial inteligence is. Watson.2 promises to be very interesting.
- Joined
- 14 Mar 04
- Moves
- 213509
Originally posted by coquetteMarketing, plain and simple.
There is a part of all this that is just plain gimmicky. I'm referring to the dramatic sales campaign aspect. IBM is getting a HUGE amount of free publicity for this. I don't mind that, it's the US free market at work. Also, it's not "free" as IBM had to develop the computer and they heavily advertized the event. I presume that they were heavy advertizers d ...[text shortened]... the art of aritificial inteligence is. Watson.2 promises to be very interesting.
Originally posted by gambit3The repeat of an incorrect answer happened during a regular jeopardy round where players ring in, IIRC. Repeating an incorrect response wouldn't happen to humans because they hear each other. Apparently the answer made by the human wasn't input to Watson as the answer was given. If it was, it would be a trivial programming effort to have Watson go to its runner-up answer or not ring in if that answer didn't score a high enough confidence rating. So it wasn't a moron, it just wasn't listening. Maybe it was dreaming of electric sheep. 🙂
Jennings rocked today but it was not enough. P.S. The computer made two moronic mistakes in the match. One it repeated a incorrect question and two it called a Canadian city a U.S. city. I do not believe a person would ever make the first mistake.
Toronto? Maybe a categorization error (e.g., listed under North American cities or just airports.) But it only had a confidence rating of 14% compared to Chicago's 11%. But being Final Jeopardy, giving an answer with a low confidence rating is better than not answering.
Originally posted by gambit3did they type the answers into it or does it perform voice recognition?
Jennings rocked today but it was not enough. P.S. The computer made two moronic mistakes in the match. One it repeated a incorrect question and two it called a Canadian city a U.S. city. I do not believe a person would ever make the first mistake.
uOriginally posted by SuzianneYep, I meant selection of the next question when they answer correctly, not the categories before the game starts. I guess I should have worded that one better.
I'm pretty sure sonhouse meant choosing which of the six categories presented to get an answer from during the game. I don't want to put words in his mouth or speak for him, but this seems pretty obvious to me. I don't see how you could think he meant IBM selecting which categories would be in the game.
Originally posted by coquetteDuring the show they showed how previous versions of Watson fared against jeopardy players, there was a lot of reworking of the computer to arrive at this version. Even this version was not 100%, it made stupid errors also. It was the first day that overwhelmed the humans and allowed it to win big time but Jennings stepped up the level on the final day but not enough to overcome the first days devastating loss.
There is a part of all this that is just plain gimmicky. I'm referring to the dramatic sales campaign aspect. IBM is getting a HUGE amount of free publicity for this. I don't mind that, it's the US free market at work. Also, it's not "free" as IBM had to develop the computer and they heavily advertized the event. I presume that they were heavy advertizers d ...[text shortened]... the art of aritificial inteligence is. Watson.2 promises to be very interesting.
Originally posted by zeeblebot"When Watson receives a Jeopardy answer (as an electronic text file) it parses the language to come up with at least three possible responses."
did they type the answers into it or does it perform voice recognition?
Reference:
http://news.yahoo.com/s/pcworld/20110215/tc_pcworld/ibmswatsontiesforjeopardyleadvictorynotcertain_1
It's well explained there.