• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

Inverse inverse Turing test

MRC_Hans

Penultimate Amazing
Joined
Aug 28, 2002
Messages
24,961
The classic inverse turing test is, of course, a human sucessfully impersonating a machine. (Against a qualified opponent, that is not necessarily easy, except for consistent '>unknown command' replies), but I'm currently in the process of a double inverse turing test, where the purpose is to see if a certain human is cabable of realizing he is interacting with a program. (So what inverses it is that the human is under test, the program is rather obvious for most)

Of course, this can be said to be off topic in the actual threads, but I hope the moderators will accept this for two reasons:

1) The OP has declared the topic closed (but continues to post in the thread).

2) It is an otherwise interesting and relevant experiment.

Some have speculated that the user in question is in fact a program and not a human, which would make it a double Turing test. :p

Hans
 
I think I see where you are going with this... Should be fun! :)
 
Perhaps, having been accustomed to communicate with a non-automation under my humble alias (I hope!), we might hardly expect the testee to realize the full ramifications, but I at least aspire for the equivalent of "hey, you're pulling my leg!".

(Intentionally taxing language style to deny subject comprehension)

Hans
 
See, and here I thought the inverse would be to ask a computer if it is dealing with a human or computer. Now that would be interesting.
 
I've shouted at many a stupid machine... many of my computers do what I tell them to do, not what I want them do!
And I've worked with some guys that simple machines like screwdrivers could tax their hand-eye-brain connection.
 
See, and here I thought the inverse would be to ask a computer if it is dealing with a human or computer. Now that would be interesting.

Mmmm, in fact I think that would be possible. The algorithm would extimate the level of determinism in the the replies. Computers are always deterministic (even when executing random algorithms).

Hans
 
Interesting idea, I can think of posters who are not likely to notice. :D
 
Mmmm, in fact I think that would be possible. The algorithm would extimate the level of determinism in the the replies. Computers are always deterministic (even when executing random algorithms).
I'm not sure what you mean by testing the level of determinism.

Humans make for poor random number generators, so testing for randomness is out. Computers can generate pseudo random numbers that are nearly impossible to pragmatically compute, so trying to find the connection between numbers is out.

If the human's trying to act human, test for something humans can do easily that computers cannot yet--and you basically have a captcha.
 
Last edited:
I'm not sure what you mean by testing the level of determinism.

I would look for repetitive patterns. If there are repetitive patterns with a narrow normal distribution (on the level of noise), the source is likely to be a program.

Humans make for poor random number generators, so testing for randomness is out. Computers can generate pseudo random numbers that are nearly impossible to pragmatically compute, so trying to find the connection between numbers is out.

If random parameters are well distributed, that is entirely random, or have a perfect normal distribution, they are almost certainly computer generated.

For instance, if you were to map the occurrence of certan words or word combinations in the random philosophy generator, you would very likely find that they are highly systematic.

If the human's trying to act human, test for something humans can do easily that computers cannot yet--and you basically have a captcha.

How will a computer test for something computers cannot do?

ETA, OK looked up captcha. Yes, that might be a possibility.

Hans
 
Last edited:

Back
Top Bottom