Outsmarting the worlds smartest computer

[MetalAtlas]

So, I don't know how many of you were paying attention, but IBM created a computer to specifically compete against the best jeopardy players, where it destroyed on the first day. However, it answered the Final Jeopardy question wrong, of which i answered correctly. How many times do you get to say you outsmarted the worlds smartest computer? Anyone else watch jeopardy tonight?

[potter_rocks_out_loud]

Nice.

[Coolboyrulez0]

I did watch it yesterday and today. (& will tomorrow)

I most genuinely enjoyed the wagers of Watson and some of the other possible answer choices it proposed.

Best part imo was the daily double where it was only 32% sure the answer is valid, and still went for it and it was correct. Exciting stuff.

[Izzy]

I've heard about it. Very impressive technology.

[Duplica]

Originally Posted by potter_rocks_out_loud

Nice.

Ken Jennings is a hero haha

[MrGiggles]

"Ken"
"1920's"
"nope. Watson?"
"1920's"

lmfao

[fido123]

Only as smart as the programmer who programmed it in the end but regardless pretty impressive.

[ninjaKIWI]

Originally Posted by fido123

Only as smart as the programmer who programmed it in the end but regardless pretty impressive.

Well, no, not exactly. What it does is it searches through a colossal library and logically decides on the best answer.

[fido123]

And who programmed the logic to decide on the best answer?

[ninjaKIWI]

Originally Posted by fido123

And who programmed the logic to decide on the best answer?

It used machine learning to teach itself, along with a hardcoded rules. It's constantly teaching itself things about the game as it plays, based on the correct answers that others give. I guess I see what you mean, though.

[fido123]

Originally Posted by ninjaKIWI

It used machine learning to teach itself, along with a hardcoded rules. It's constantly teaching itself things about the game as it plays, based on the correct answers that others give. I guess I see what you mean, though.

In the end it's still all programming. People taught it how to think, and if they taught it how to think perfectly it would probably get every answer perfect. I'm not saying the programming would give out the same questions in a game of Jepardey, but the better the algorithms it uses made by the programmer, the more accurate it will be.

[MetalAtlas]

Originally Posted by fido123

In the end it's still all programming. People taught it how to think, and if they taught it how to think perfectly it would probably get every answer perfect. I'm not saying the programming would give out the same questions in a game of Jeopardy, but the better the algorithms it uses made by the programmer, the more accurate it will be.

yeah but it wasn't just one person, it was MANY people. So if you're playing jeopardy, and there are 15 people behind the podium, that are all hooked up so they can read each other's thoughts, each with a signal device, then yeah its the same lol

[Glucuronolactone]

I didn't watch any this week but I read an article about it. I found it pretty amazing. I think they imported around 20 million pages of information into the computer. The computer is also able to figure out the answer with clues, as mentioned before. I think the example given in the article was the homonym of bark. The bark of a a tree is not the same as a dog's bark. Anyway, here's the article if anyone's interested:
http://money.cnn.com/2011/01/13/tech...tson/index.htm

[Reach]

I watched it. I don't particularly like Jeopardy as a show, but I was genuinely interested in this spectacle. I follow the chess scene pretty closely, and I still remember the monumental shift in chess AI after Deep Blue beat Kasparov. It was just as exciting to see a computer tackle the problem of language.

I mean, it's obvious the computer can store and retrieve indefinitely more information than any human opponent could, but actually understanding what has been asked and finding the answer is another problem entirely.

Ken and Brad got completely dominated, for the most part. I mean, the second half was brutal. They basically just stood there and took it from Watson, though I did expect as much.

Challenges on the horizon? Computers that can beat the best Go players, or even poker players.

[Izzy]

Originally Posted by fido123

In the end it's still all programming. People taught it how to think, and if they taught it how to think perfectly it would probably get every answer perfect. I'm not saying the programming would give out the same questions in a game of Jepardey, but the better the algorithms it uses made by the programmer, the more accurate it will be.

It does have close to perfect programming. It can't find the answer instantly because its searching through databases of billions and billions of things. Even with the most efficient algorithms humans can still occasionally out preform it. It just shows how amazing the human brain is at recalling a specific piece of information at will.

[Rubin0]

I can't even outsmart my own dog. Props to these people if they can even outsmart this thing even once.

[Crashfan3]

They probably found some old couple in Mississauga who had every episode of Jeopardy on tape, and had the robot watch every single tape.

Poor robot.

[justin_ator]

I don't recall the amount right now, but doesn't it search through several terabytes of information?

[ninjaKIWI]

Originally Posted by justin_ator

I don't recall the amount right now, but doesn't it search through several terabytes of information?

It has 15TB of RAM, but the information it searches through is actually under 1TB.

[aperson]

Originally Posted by Reach

Challenges on the horizon? Computers that can beat the best Go players, or even poker players.

Limit? Done. No limit? Probably at least 20 years.
Go? I have no idea but the combinatorial complexity on that makes me think it's also a pretty big longshot.