Support NeoGAF

[pj]

...and how is it fair for the computer to be able to buzz in? If they wanted, they could have the thing instantly buzz in everytime so that if it thought it knew the answer, it'd definitely get the buzz in.

The article says a human player gets the answer right when they buzz in 85% of the time, and the computer isn't that good yet, so if it buzzed in every time it would be wrong a lot more. Also I'm sure it takes time for it to come up with an answer, or at least they would put a sleep(1000) in there to make it more fair.

 

[Diablohead]

The next step is Skynet

 

[koam]

This is EASIER to program than chess. What's the big deal?

 

[ChoklitReign]

Oh oh..
I think that they (robots) have made the first attack now..

From a swedish newspaper:


Robot misshandlade anställd
http://www.aftonbladet.se/nyheter/article5004310.ab

Företagets robot gick oväntat till attack och skadade en anställd allvarligt.

Nu har företaget fått böta 25 000 kronor som ansvarigt för händelsen.

– Jag har aldrig hört talas om en robot som misshandlat en människa på det här sättet, säger åklagare Leif Johansson till TT.
--------

Translated:

Robot assaulted employee

The company robot unexpectedly attacked and injured severley, an employee.
The company must now pay 25 000 SEK for this event.

"I have never heard about a robot assaulting a human being in this way, said prosecutioner Leif Johansson"
----------

it has begun..

It's happened several times before in Japan. One of them died.

 

[Zoramon089]

I don't see why this would be such a huge advancement? I mean, it'd simply be taking in the category title and square "question", searching on Google, parsing the results and returning it with a "What is..." before the answer

 

[Saargabath]

I lol'd :lol

 

[JoshuaJSlone]

This is EASIER to program than chess. What's the big deal?

People watch Jeopardy.

 

[Evlar]

Prediction for the next breakthrough: a computer with a camera that can view moving images just like the human eye!

 

[iapetus]

This.

Whats hilarious is that even if the day comes where computers can completely master Go, all we have to do is say, "okay, time to make 21x21 boards."

No, as any computer algorithm to master Go is unlikely to use exhaustive search (in the same way that the best Bridge computer players don't - though the same approaches wouldn't work, obviously...)

 

[AlphaTwo00]

This is EASIER to program than chess. What's the big deal?

:lol

I hope you are not serious.

:lol

 

[AlStrong]

Spoiler

Sup dawg, we heard you liked to smile, so we put the smile in the smile so you can smile while you smile.

 

[t4ng0]

I agree that they could attempt to program for common templates, however 99% of the time, the category names don't describe what the ruleset is, so the computer would have no way to know how it is supposed to interpret those categories.

Alex verbally describes what the theme of the category is to the contestants. This AI would have to pick up what is said through microphones and interpret the audio into a ruleset.

It's just not going to happen unless Jeopardy specifically avoids these types of things to accommodate the AI's programming.

Thats the point of the program - to deal with those subtleties. Constructing a knowledge base is trivial, understanding what to do with/how to interpret is everything.

Also, chess has not been "solved."

 

[LakeEarth]

That aside, I have to imagine Jeopardy would have to tailor it's questions to be in AI understandable form. When they have random categories or video clues, wtf is this thing supposed to do?

Aside from video clues, random categories is EXACTLY what this program is all about. To figure out what the category and question means and answer correctly. That's the challenge.

The article says a human player gets the answer right when they buzz in 85% of the time, and the computer isn't that good yet, so if it buzzed in every time it would be wrong a lot more. Also I'm sure it takes time for it to come up with an answer, or at least they would put a sleep(1000) in there to make it more fair.

Chances getting the program to understand the question and query the answer fast enough to beat a human is another one of the goals of this program. If this program is actually going to face Jeopardy contestants, it would also have a "% certain" calculation to help it decide if it should answer or not.

 

[Keylime]

Aside from video clues, random categories is EXACTLY what this program is all about. To figure out what the category and question means and answer correctly. That's the challenge.

Unless the thing is able to listen to what Trebek says is the point of the category in those oddball cases, it's not going to be able to be successful.

 

[LakeEarth]

Unless the thing is able to listen to what Trebek says is the point of the category in those oddball cases, it's not going to be able to be successful.

I would hope that for the categories that Alex has to explain, the explanation is included for the program to understand as well. I dunno, the article doesn't mention that.

 

[MisterHero]

but he did ask it in the form of a question.

note the question mark

lawlz anyways