Hard Light Productions Forums
Off-Topic Discussion => General Discussion => Topic started by: Black Wolf on June 25, 2008, 09:05:20 am
-
I was thinking about this the other day - do you think that a computer could pass the Turing test without hugely sophisticated programming if it simply had enough responses to draw from? Something on a similar scale to wikipedia - thousands of people throwing up theoretical questions (possibly in a sort of "What is the X with your Y?" sort of system) and answers. Have a human team vet the questions and responses to eliminate the absolutely ridiculous and then wait until it reaches critical mass. If you had even a percentage point of wikipedia's participation, you could have a mammoth amount of data - and better yet truly human data. From such a source material base, even relatively simple programming should be able to get a computer through the test, no? Or am I on the wrong track here?
Sure, it wouldn't exactly be in the spirit of the test, but it'd be... interesting :)
-
Theoretically, it's entirely possible. However, the Turing test hinges on being able to perfectly mimic the responses you would expect from a human being, and having a machine simply draw responses from a pool of preprogrammed answers just wouldn't be as smooth as if there were an actual intelligence. Of course, what is human speech if not drawing responses from our own little internal wikis? I don't really know enough about computing to give an informed opinion on this, but you're definitely on to something here. Cheap trick, but definitely something.
What'd really amaze me is if the machine you propose goes on to get the extra credit in the Turing test: Not only convincing the operator that he's talking to another person, but convincing the operator that he's a machine.
-
What'd really amaze me is if the machine you propose goes on to get the extra credit in the Turing test: Not only convincing the operator that he's talking to another person, but convincing the operator that he's a machine.
That's when world war 3 starts.
-
What'd really amaze me is if the machine you propose goes on to get the extra credit in the Turing test: Not only convincing the operator that he's talking to another person, but convincing the operator that he's a machine.
A machine that will get stoned with us? ****in' righteous.
-
I don't think it would work.
You'd need to have a personal history where things associate without contradiction. How would that succeed with brute force/statistical methods?