IBM Watson

[CaptainGiblets]

I have just heard about the computer and was amazed with it's intelligence. If you don't know what Watson is, here is the page for the computer http://www.ibm.com/smarterplanet/us/en/ibmwatson/. Also just random is that it played on the game show "Jeopardy" and won by lots of money (Not sure how much)

Edit: Found the video

Edited May 24, 2014 by CaptainGiblets

[NFUN]

It's a pretty cool and advanced computer, but it still can't pass the Turing Test AFAIK (we can argue about the test's efficacy elsewhere) and it isn't *that* much more intelligent than other programs like it. It is mostly known for its ability to make relevant connections, which is why it can beat Jeopardy while other AI's can't (they miscontrue what the question/statement is asking) and why it is being targeted for medical diagnoses.

I may or may not have pulled the 'connections' part of my post out of my ass/from potentially false logic

[phoenix_ca]

...why it is being targeted for medical diagnoses.

I'm amazed that IBM made such a huge mistake in reasoning on the page discussing that.

According to one expert, only 20 percent of the knowledge physicians use to diagnose and treat patients today is evidence based. Which means that one in five diagnoses is incorrect or incomplete.

Completely, utterly wrong. (Not because the figure is wrong; it might be right.)

Edited May 24, 2014 by phoenix_ca

[Woopert]

I've seen it before on TV. My guess is that it interprets the question somehow, and searches the whole web (or as much as it can) for an answer.

[phoenix_ca]

I've seen it before on TV. My guess is that it interprets the question somehow, and searches the whole web (or as much as it can) for an answer.

Then your guess is wrong. When it was demonstrated on Jeopardy, it queried its own database, which pulled from print sources (and a whole lot of other stuff) too. Though it didn't work very well until they used machine learning to teach it what questions actually meant. No internet connection required, the entire thing was self-contained. This is clearly explained in the video linked in the OP.

Edited May 24, 2014 by phoenix_ca

[zxczxczbfg]

... No internet connection required, the entire thing was self-contained ...

Indeed, it was in fact specifically not connected to the internet in order to avoid cheating during the competition.

[50ck37]

It is indeed impressive what a machine can do nowadays just by having huge amounts of data available for querying (and, of course, being able to identify a question).

[YNM]

It is indeed impressive what a machine can do nowadays just by having huge amounts of data available for querying (and, of course, being able to identify a question).

... Aand that's what Google want to do with all the data it already (sneakily) get.

[50ck37]

... Aand that's what Google want to do with all the data it already (sneakily) get.

Yes, they want to organize it and make it freelly available to everyone, as they say. I don't really trust that, as with great databases come great responsabilities and with great power comes great greed.

[vger]

It's getting an upgrade too now. Evidently Watson's next task is to render internet politicking obsolete. They expect it to be capable of debate, analyzing a person's argument and submitting a counter-argument, backing it up of course with a thousand examples from its near infinite memory.

Personally I want to see a variant of that, where Watson sits in on election debates, and points out whenever a politician contradicts something they have said in the past.

Sadly, when the technology gets that good, what are the odds that someone won't slip a little bias into the machine?

[Dkmdlb]

Personally I want to see a variant of that, where Watson sits in on election debates, and points out whenever a politician contradicts something they have said in the past.

What's the use in that? You might as well have it point out whenever it sees the effects of gravity. Why not put it to work on a problem we don't already know the answer to? Like, I don't know, arguing on forums or something.

[vger]

What's the use in that? You might as well have it point out whenever it sees the effects of gravity. Why not put it to work on a problem we don't already know the answer to? Like, I don't know, arguing on forums or something.

WE know the answer. A lot of people still don't, otherwise nobody would bother with public debates anymore.

The whole point of having an automated 'referee' would be to force honesty. Watson could catch their lies any time they do it. Most of the time, it soars over people's heads, because nobody can be bothered to know every single thing a politician has said during a campaign. For Watson, that would be a cakewalk.

[phoenix_ca]

Why not put it to work on a problem we don't already know the answer to? Like, I don't know, arguing on forums or something.

Watson isn't able to make conjecture, or to perform the creative and problem-solving tasks that humans can do so as to solve a problem that has no known solution yet. It is instead querying a massive database of information generated by humans. As long as someone has already made the connection between say, Coffee-Mate and "friend", Watson can find that connection and make it itself, but as far as I'm aware, it's unable to theorise. If it could, we'd have the proverbial Star Trek computer.

[50ck37]

Watson isn't able to make conjecture, or to perform the creative and problem-solving tasks that humans can do so as to solve a problem that has no known solution yet. It is instead querying a massive database of information generated by humans. As long as someone has already made the connection between say, Coffee-Mate and "friend", Watson can find that connection and make it itself, but as far as I'm aware, it's unable to theorise. If it could, we'd have the proverbial Star Trek computer.

There's this one that is really interesting: http://en.wikipedia.org/wiki/Robot_Scientist

[Ralathon]

What's the use in that? You might as well have it point out whenever it sees the effects of gravity. Why not put it to work on a problem we don't already know the answer to? Like, I don't know, arguing on forums or something.

I can already see the future. A forum with a couple dozen Watson's as members, each with different databases to induce some differences in their way of thinking. Internet goers can post a new thread with some argument/flamebait and the different Watsons start arguing among themselves while the humans eat popcorn.

I'd totally pay to watch that.

[WinkAllKerb'']

this is just one more 'specific' attempt to reproduce some of the existing neuronal scheme. If you seek more info regarding neuronal scheme and computing reproduce attempt you could check both "brain" research big chiefs annual reports here and there (no fresh link or name to blame my side sry find them if you can by luck, i hope they fixed the ability to find/access them in accordance with there own disclaimer due to there lack with computer language or lowfounding to hire someone to do it => no student + high% of there own colleague: anecdotic roflmIao but can't resist sry) and computer software history.

You might get some clue sherlock.

[vger]

I can already see the future. A forum with a couple dozen Watson's as members, each with different databases to induce some differences in their way of thinking. Internet goers can post a new thread with some argument/flamebait and the different Watsons start arguing among themselves while the humans eat popcorn.

I'd totally pay to watch that.

They will have to be taught to mimic the personalities of internet trolls too, so they can obey Godwin's law, make derogatory comments about mothers, and accuse each other of still living in the server room of their Creator and never getting laid.

Edited May 26, 2014 by vger

[WinkAllKerb'']

They will have to be taught to mimic the personalities of internet trolls too, so they can obey Godwin's law, make derogatory comments about mothers, and accuse each other of still living in the server room of their Creator and never getting laid.

yup this would give fun answer like: "don't break a mirror or you'll get cursed ..." sad so sad, i tend to prefer chips vs popcorn but's this is very subjective ...

[50ck37]

I wonder what kind of game threads these robots would start...

for instance:

- Kerbal System/ASCII art - only alpha characters allowed

- Hack the previous poster's code: make it overflow somehow

Edited May 27, 2014 by 50ck37

[rkman]

WE know the answer. A lot of people still don't, otherwise nobody would bother with public debates anymore.

The whole point of having an automated 'referee' would be to force honesty. Watson could catch their lies any time they do it. Most of the time, it soars over people's heads, because nobody can be bothered to know every single thing a politician has said during a campaign. For Watson, that would be a cakewalk.

This being a machine made by humans, and humans being known to make errors, how sure can we be this automated referee would be correct in its judgement?

Edited May 29, 2014 by rkman

[Rakaydos]

This being a machine made by humans, and humans being known to make errors, how sure can we be this automated referee would be correct in its judgement?

What judgement? it would simply be putting canidates on the spot. "Senator, during your presidential campain you've pledged to make sure schools, police, and fire departments are well funded. How does this uphold the campain promices you made in the primaries to cut the budget "by any means possible?"

[WinkAllKerb'']

In any cases , the more case your paying attention at, the more case you will need to case the case. Notice that some case dislike to get out of there case and that some case are not always related to others cases.

EDIT:

it work kinda like that with more finess, variant and similitude

Edited May 30, 2014 by WinkAllKerb''

[rkman]

What judgement?

The judgement of being found to be dishonest.

(the statement to which i replied: "The whole point of having an automated 'referee' would be to force honesty")

it would simply be putting canidates on the spot. "Senator, during your presidential campain you've pledged to make sure schools, police, and fire departments are well funded.

How does this uphold the campain promices you made in the primaries to cut the budget "by any means possible?"

Don't need fancy AI to do that. Journalists do that all the time, and it achieves very little. People vote for that politician next time around just as well.

[Javster]

It's a pretty cool and advanced computer, but it still can't pass the Turing Test AFAIK (we can argue about the test's efficacy elsewhere) and it isn't *that* much more intelligent than other programs like it. It is mostly known for its ability to make relevant connections, which is why it can beat Jeopardy while other AI's can't (they miscontrue what the question/statement is asking) and why it is being targeted for medical diagnoses.

I may or may not have pulled the 'connections' part of my post out of my ass/from potentially false logic

Not to contradict you, but a human's result is not 100%

[vger]

Don't need fancy AI to do that. Journalists do that all the time, and it achieves very little. People vote for that politician next time around just as well.

There's a big difference between a Journalist doing hours of research on a candidate AFTER a debate is already over.

Watson could pull up the information immediately. What I'm talking about would happen LIVE, during the debate, and broadcast for the whole country to see. Effectively, Watson would serve as a co-moderator, and could directly address a candidate when it caught them in a lie.