Tuckins1
Ideal_Rock
- Joined
- Apr 13, 2008
- Messages
- 8,614
slg47|1297794823|2852470 said:Laila-tomorrow is double and final jeopardy, and then the next day is a full game
davi_el_mejor|1297815017|2852737 said:Watson, an alternate spelling of Skynet![]()
stephb0lt|1297956329|2853848 said:I watched all three days. While it is impressive that the computer is able to generate the correct answer so frequently, on some of the questions it was way off base, which shows there are definite gaps in the language recognition. Overall, I'm not that overwhelmed by Watson's victory. In my opinion, what that comes down to is Watson being able to accurately time his button pressing based on the signal it is getting that the clue has been read. When very good Jeopardy players play each other, it comes down to the button skills. There were many questions that Watson got that Ken and Brad both knew the answer to, I would wager. But a computer isn't going to buzz in early and lock itself out, or buzz in a split second too late. So in the event that Watson knew the answer, I would say that he had the buzzer advantage at least 90% of the time, if not more.
Still, it was a fun experiment to watch.
Elrohwen|1297956858|2853856 said:stephb0lt|1297956329|2853848 said:I watched all three days. While it is impressive that the computer is able to generate the correct answer so frequently, on some of the questions it was way off base, which shows there are definite gaps in the language recognition. Overall, I'm not that overwhelmed by Watson's victory. In my opinion, what that comes down to is Watson being able to accurately time his button pressing based on the signal it is getting that the clue has been read. When very good Jeopardy players play each other, it comes down to the button skills. There were many questions that Watson got that Ken and Brad both knew the answer to, I would wager. But a computer isn't going to buzz in early and lock itself out, or buzz in a split second too late. So in the event that Watson knew the answer, I would say that he had the buzzer advantage at least 90% of the time, if not more.
Still, it was a fun experiment to watch.
I've heard this a lot, and I do agree that there were many questions that Ken and Brad knew, but they weren't as fast to buzz in as Watson. Watson's buzzer speed was the key reason he won by so great of a margin.
Still, I think that overlooks how truly impressive it is to have a computer able to understand human language with the level of complexity Jeopardy uses. Sure it's just a very advanced search engine, but it really is a huge advancement from where our current search technology is. Watson also learns as he goes and modifies his future answers based on what he's gotten wrong.
It took an IBM team 4 years to do this - I think it's more impressive than just buzzer speed.
Personally, I'd like to see Watson compete with Tournament level questions (the ones used this week seemed to be at a normal Jeopardy difficulty) with a slower buzzer speed that matches human abilities. I don't think Watson would blow away his competition, but I still think he would win.
stephb0lt|1297961262|2853919 said:Elrohwen|1297956858|2853856 said:stephb0lt|1297956329|2853848 said:I watched all three days. While it is impressive that the computer is able to generate the correct answer so frequently, on some of the questions it was way off base, which shows there are definite gaps in the language recognition. Overall, I'm not that overwhelmed by Watson's victory. In my opinion, what that comes down to is Watson being able to accurately time his button pressing based on the signal it is getting that the clue has been read. When very good Jeopardy players play each other, it comes down to the button skills. There were many questions that Watson got that Ken and Brad both knew the answer to, I would wager. But a computer isn't going to buzz in early and lock itself out, or buzz in a split second too late. So in the event that Watson knew the answer, I would say that he had the buzzer advantage at least 90% of the time, if not more.
Still, it was a fun experiment to watch.
I've heard this a lot, and I do agree that there were many questions that Ken and Brad knew, but they weren't as fast to buzz in as Watson. Watson's buzzer speed was the key reason he won by so great of a margin.
Still, I think that overlooks how truly impressive it is to have a computer able to understand human language with the level of complexity Jeopardy uses. Sure it's just a very advanced search engine, but it really is a huge advancement from where our current search technology is. Watson also learns as he goes and modifies his future answers based on what he's gotten wrong.
It took an IBM team 4 years to do this - I think it's more impressive than just buzzer speed.
Personally, I'd like to see Watson compete with Tournament level questions (the ones used this week seemed to be at a normal Jeopardy difficulty) with a slower buzzer speed that matches human abilities. I don't think Watson would blow away his competition, but I still think he would win.
I don't disagree with you that the system they built was very impressive - it was the first thing I said. I just felt like for the purposes of competition the design was flawed, but it's hard to fix it because a computer can't have nerves to affect its reaction speed. I guess it was fed some sort of electronic signal when the clue was finished being read so for that purpose it was probably impossible for it to buzz in early as well.
I would love to be able to see more about the algorithms the system is using to parse the clues - it seemed the style of the question really affected how the system responded. One example I can think of from last night is the "Actors who Directed" (not the right name, but hopefully you know what I mean) category. The clues were only the name of the movie(s). Watson seemed to have trouble taking that info alone and coming up with an answer. Same with the Harry Potter clue from day 1 - there was almost too much info in the clue in that case so Watson came up with the most likely answer based on all the terms in the question.
kenny|1297967560|2854023 said:I think the disadvantage humans have is the relatively long time it takes to move the muscles to push the button, not the electrical impulses moving more slowly in human nerves than computer wires.
Elrohwen|1297967984|2854029 said:kenny|1297967560|2854023 said:I think the disadvantage humans have is the relatively long time it takes to move the muscles to push the button, not the electrical impulses moving more slowly in human nerves than computer wires.
I think Steph meant that the computer wouldn't be nervous which could effect reaction speed of a human competitor - not that the physical nerves themselves are slowing people down.
kenny|1297993905|2854396 said:
Clairitek|1298049824|2854760 said:I was impressed with the voice recognition software that IBM developed and Watson's ability to comb through his database to find the most appropriate answer. I did, however, get pretty bored after about 2 minutes watching him school the human competition. I did feel that the questions were more simple than usual (either that or I've learned a lot since I last watched Jeopardy). I suppose since I am not into computers, programming, or understanding how it all works it didn't blow me away because I don't get all of the intricacies of making a system like that and what makes it special.