THIAHM: Turing Method - A definition of self awareness

[Brad Theado]

In 1950, the computer pioneer Alan Turing posed a definition of determining intelligence. In order to tell whether a machine can or cannot be considered intelligent as a human, he proposed the famous Turing test. Two keyboards, one connected to a computer, the other leads to a person. An examiner types in questions on any topic he likes. Both the computer and the human type back responses that the examiner reads on the respective computer screens. If he cannot reliably determine which was the person and which the machine, then we say the machine has passed the Turing test.

Is this enough of a test for self awareness? In our book, "Mike" is declared to be self aware by his ability to do just what Turing describes. It seems to me that tricking me into thinking that you have answered all of my questions is not the same thing as being self aware. What do you think would be a better test of self awareness? How scared are you of thinking that a computer could reach self awareness?

[Sean]

There are two problems with the Turing Test:

1) People were fooled by ELIZA.

2) I've encountered several people on the Internet that I'm not certain aren't A.I. experiments.

(There's also the idea of the Chinese Room. I don't buy it, but it is a serious challenge to the idea of A.I.)

[Patrick]

I think the problem with the Turing test is that it defines intelligence in terms of input and output but doesn't say a lot about what goes on in between. I can sit quietly thinking about things without ever giving any outward indication of the mental activity.

[Mnchur]

"I can sit quietly thinking about things without ever giving any outward indication of the mental activity. "

this is very true but as the name states the turning test is at its heart a test.

A test needs inputs. Which means, whatever is being tested needs to produce some sort of measurable output. Making a test for something with no output would be pretty useless.

A turning test should therefore only be used when the "intelligence" in question produces a measurable output of some sort.

[Patrick]

Mnchur wrote: "...the turning test is at its heart a test."

True but the original poster asks if this is enough of a test for self awareness. However if you ask that question then how can you be sure that anything you communicate with human or computer is self aware.

My main problem with the Turing Test and AI research is that the test is often used as a definition. I agree that good software engineering processes need test cases. If you have a black box and if you have a well bounded set of inputs and outputs then maybe you have enough to define what the black box you are testing is and how it works. But what are the bounds of intelligent and self aware input and output? I would say its pretty much infinite. If I can't enumerate the inputs and outputs of the test then how can I use the test as a definition?

[Stan]

I remember reading years ago about one of the Turing Test contest winners.

The winner, who convinced the most people that his AI was a human did it by making his program actually appear to be stupider than everyone elses.

Introducing typing mistakes. Answering some questions in-correctly. Pausing before answering. All these made his program look more human to the people in the other room, because they all assumed that a real computer wouldn't make these kind of mistakes.

It appears that to make something to appear more human all you have to do is to make it be less perfect.

[Jenny (Reading Envy)]

Oh dang I had copied and pasted an entire conversation I had with ask.com, just to see if it even came close to a real conversation (it doesn't).

[Beth]

How to test a computer for self-awareness:

1. Turn off computer.
2. Place a small dot of odorless, tasteless paint on the monitor where the computer cannot see it (blue is a common choice, but any color will work).
3. Put a mirror in the computer's line of sight.
4. Turn on the computer.

If the computer, after some exposure to the mirror, examines the dot in the mirror then touches the dot on their forehead, we have good evidence of self-awareness.

The evidence for self-awareness gets stronger if the computer continues interacting with the mirror by looking at and examining parts of its body it cannot otherwise see (inside the mouth, back of the head, in the ears).