Human Rights For Robots.

asim18

asim18

asim18

Soldato
Joined
5 Dec 2006
Posts
15,370
sigma said:
  • Simultaneous tests as specified by Alan Turing
  • Each judge was involved in five parallel tests - so 10 conversations
  • 30 judges took part
  • In total 300 conversations
  • In each five minutes a judge was communicating with both a human and a machine
  • Each of the five machines took part in 30 tests
  • To ensure accuracy of results, Test was independently adjudicated by Professor John Barnden, University of Birmingham, formerly head of British AI Society

http://www.reading.ac.uk/news-and-events/releases/PR583836.aspx
Click to expand...

What I'm asking is for actual evidence of this test being passed. As in a transcript of these "300 conversations" would be very useful, and extremely interesting, if you can please link. Some head of an "AI society" saying something is passing a test is not evidence of a test being passed at all! I'm sure this guy spouts a lot of BS to get funding for his little society. This is the difference between believing and knowing. I want to know if the test has been passed, not merely believe it was passed.

Also there's a blooming error in the first 2 words of that article lmao. I have no reason to take anything on that website about "language" seriously! "An historic"? Seriously? Who writes like that? Maybe if you pronounce "history" as "istory" you can get away with speaking "an istoric", but you cant do that in writing lmao :D
 
Last edited:

stewski

stewski

stewski

Soldato
Joined
12 Nov 2015
Posts
4,010
asim18 said:
What I'm asking is for actual evidence of this test being passed. As in a transcript of these "300 conversations". Some head of an "AI society" saying something is passing a test is not evidence of a test being passed at all! I'm sure this guy spouts a lot of BS to get funding for his little society.

Also there's a blooming error in the first 2 words of that article lmao. I have no reason to take anything on that website about "language" seriously! "An historic"? Seriously? Who writes like that? Maybe if you pronounce "history" as "istory" you can get away with saying "an istoric", but you cant do that in writing lmao :D
Click to expand...

You can make the case that the data should be entirely open, I'd agree.
You could even make the case that evidence/results should be repeatable, I'd agree.

Having suggested there is no natural language processing required in the IBM Watson Jeopardy win, I'd avoid the Grammar Nazism's, "I ate you enry iggins", It really doesn't add any value.
 

sigma

sigma

sigma

Soldato
Joined
13 Nov 2006
Posts
23,977
asim18 said:
What I'm asking is for actual evidence of this test being passed. As in a transcript of these "300 conversations" would be very useful, and extremely interesting, if you can please link. Some head of an "AI society" saying something is passing a test is not evidence of a test being passed at all! I'm sure this guy spouts a lot of BS to get funding for his little society. This is the difference between believing and knowing. I want to know if the test has been passed, not merely believe it was passed.

Also there's a blooming error in the first 2 words of that article lmao. I have no reason to take anything on that website about "language" seriously! "An historic"? Seriously? Who writes like that? Maybe if you pronounce "history" as "istory" you can get away with speaking "an istoric", but you cant do that in writing lmao :D
Click to expand...

I recommend that you read post #281 and argue against that instead.
 

Zethor

Zethor

Zethor

Soldato
Joined
13 Nov 2013
Posts
4,294
stewski said:
I don't care.

As in the Turing test. if a computer can imitate a human, does it matter how it works?
Click to expand...


dowie said:
I'm not sure why those are the only two options and I'm not seeing the reason for your belief that we must first be able to model biological systems before making progress in AI research.
Click to expand...


Let me put this in another perspective. Would it be possible to create a program that can beat a grand master at chess without input regarding the rules of the game? That's the reason we've made leaps with weak-AI but only baby steps towards strong-AI. We don't understand the rules of the game(yet).
 

Tefal

Tefal

Tefal

Caporegime
Joined
30 Jun 2007
Posts
68,784
Location
Wales
dowie said:
It doesn't therefore follow that we need to simulate biological systems in order to make a breakthrough.
Click to expand...

exactly, you'd never base an F1 car on any form of biological locomotion as it would get destroyed by it's completely abstract car based completion.
 

Tefal

Tefal

Tefal

Caporegime
Joined
30 Jun 2007
Posts
68,784
Location
Wales
Zethor said:
Let me put this in another perspective. Would it be possible to create a program that can beat a grand master at chess without input regarding the rules of the game?
Click to expand...

Would it be possible for for a human to win against a grand master without knowing the rules?:confused:
 

dowie

dowie

dowie

Caporegime
Joined
29 Jan 2008
Posts
58,912
Going back to the computer, yes. Not that it would add much but you could get your program to learn the rules if you want.
 

Zethor

Zethor

Zethor

Soldato
Joined
13 Nov 2013
Posts
4,294
dowie said:
It doesn't therefore follow that we need to simulate biological systems in order to make a breakthrough.
Click to expand...

Strong-AI research, by definition, is a reverse engineering process. Intelligence evolved independently in several branches of the tree of life yet it seems to be based on similar processes. There is no reason to believe there exists another form of intelligence, one which only machinery can achieve, which is what this breakthrough you mentioned seems to refer to. If you make the claim it's on you to put forward arguments, not me.

Would it be possible for for a human to win against a grand master without knowing the rules?
Click to expand...

Weak-AI has the chess rules encoded in the algorithm, that's why it's called weak or narrow, without the model it's nothing. Humans don't have a chess model written in the brain, every grand master that has ever lost did so to a human that at some point didn't know the rules.

dowie said:
Going back to the computer, yes. Not that it would add much but you could get your program to learn the rules if you want.
Click to expand...

That's not how Deep Blue or Watson work. Weak-AI needs a complete model to function it can not 'learn' the ruleas. You would need Strong-AI for that.
 
Last edited:

Tefal

Tefal

Tefal

Caporegime
Joined
30 Jun 2007
Posts
68,784
Location
Wales
ouldnt the computer being programed with the rules be exactly like a human being shown the rules?


ie

this piece can move in this pattern.


it's not like you have to work them out you're simply told them.
 

dowie

dowie

dowie

Caporegime
Joined
29 Jan 2008
Posts
58,912
Indeed, it certainly doesn't require strong AI nor does it demonstrate this supposed need to simulate biological systems. But in the example of say deep blue, if you're building a program specifically to beat people at chess I'm not sure that making the program first learn the rules is particularly beneficial.
 
You must log in or register to reply here.
Back
Top Bottom