Re: Turing Test

From: HARNAD Stevan (
Date: Tue Jun 04 1996 - 10:10:20 BST

> Date: Thu, 23 May 96 18:31:07 -0700
> From: "Dawson Jon" <>

Please use a spell-checker!

> The Turing Test (proposed by mathmatitian Turing,1950, years before the
> ability to simulate human behaviour was available) was suggested as a
> method to tell whether a computer is thinking. It is useful in the
> feild of artificial inteligence (the feild in which computer programs
> are devised to display intelegence in solving problems/producing
> conversation). A subfeild of this is computer simulation, where the
> programs try to simulate human behavior. Computer simulation may not be
> the best way of running a program, but it is supossed to simulate human
> thought processes.

Not sure what the last sentence meant: that when you are both trying to
get a computer to do something AND to do it the same way people do, the
double constraint may not lead to the best programme for the task
itself? If so, that's true, but kid-sib would have no way of knowing it,
so you could not be given credit for it in the marking.

> The Turing Test suggests that if you put two teletypes into a room,
> with a person to act as judge, who may ask any questions they like by
> typing. One teletype is connected to a person in another room, and the
> other is connected to a computer. If the judge cannot tell which is the
> 'human teletype' and which is the 'computer teletype' then the computer
> must be thinking.

First, how long should this test go on? Is 3 minutes enough?

Second, Turing did not say the computer MUST be thinking if it passes.
The point is just that you have no better (or worse) reason to revise
your conclusion that it was thinking once you were told it was a

> This idea produces in interesting logic flaw when
> considering COMPUTER SIMULATION, because just having the same
> behavioral output as a human does not mean that the computer uses the
> same cognitive processes

But this is a Granny objection unless you have a way of knowing what the
RIGHT processes are; for if you don't know which the right ones are, how
can you know which the wrong ones are?

The reply was too narrow. Go more deeply into the Turing Test (timing,
lifelong, pen-pal vs. robot version) and the broader issues, such as the
other-minds problem and reverse engineering the mind.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:43 GMT