Turing test

[Lothian]

I picked this up from Reuters

The Loenner prize committee are sorry to report that the latest Turing test applicant has failed. For a while it was looking like the machine was making rational posts indistinguishable from a human.

Unfortunately over time and on fuller analysis this deteriorated. The first noticeable thing was the repeated inappropriate use of a smilie . It then became apparent that the posts were simply meaningless waffle. Independently, sentences were considered to be vaguely addressing the subject and could possibly be seen to be human in origin. However, collectively they are judged nonsense and could only have come from an unthinking processing unit.

However this has been quite a successful experiment and it is hoped that the improved version ‘Iacchus 2’ will fare better.

 

[MRC_Hans]

LMAO!

On the Inverse Turing Challenge (ITC), however, there is a sizeable number of contenders struggling to qulaify for the quarterfinals.

The ITC awards one million homeopathic dollars for the real person who can consistently emulate an answering bot.

Hans

 

[pupdog]

It then became apparent that the posts were simply meaningless waffle. Independently, sentences were considered to be vaguely addressing the subject and could possibly be seen to be human in origin. However, collectively they are judged nonsense and could only have come from an unthinking processing unit.

Are you sure this wasn't an interview with Bush?

 

[Scot C. Trypal]

Seriously though, I wonder how limited one could make the input and reply options before a computer could consistently pass the Turing test.

If there is only one response option for computer or human, say “yes”, then it will always be indistinguishable and will only fail by a 50/50 guess. But how many options can the computer have before it’s found out?

Say the human can only input one number at a time and the computer could only respond similarly. Could I write a program which responds to a number with a number in a fashion indistinguishable from another human? I think so but it’d be tricky to get that right balance of memory, randomness, and cause.

Anyway, it would be a fun thread for the puzzle section of the JREF (Note: need nerdy sense of fun).

 

[c4ts]

I thought the laccus AI performed admirably, compared to the post generation programs modeled after Christian fundamentalists. Oh, wait, those were actual posters.