Turing test was created to test machines exhibiting behavior equivalent or indistinguishable from that of a human. Is that the sufficient condition of intelligence?
$\begingroup$
$\endgroup$
0
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
4
We don't know.
However, an important line will have been crossed - it will be impossible to tell the difference between an intelligent agent and the machine by use of a text interface. Which is the main point of the test - "if it quacks like a duck".
It is also an important philosophical point. Whether intelligence is defined purely by behaviour in an environment, or by the mechanisms that arrive at that behaviour. A suitably large database of conversational openers and "correct" responses can in theory mimic a lot of real world conversations. Some chatbots take advantage of this and use modern computer capacity to store a lot of responses, and that approach has gained competitive scores in the Loebner prize competition (although not to the stage of actually passing the test). This leads us to the Chinese Room issue, and wondering which part of the system is actually intelligent, or even how much of human conversation is actually intelligent or meaningful (and it what ways).
answered Aug 16, 2018 at 8:24
-
$\begingroup$ I don't think that this answer is correct. You should have answered "I don't know", not "We don't know". The answer to the question depends on your definition of intelligence. $\endgroup$– nbroJun 27, 2019 at 20:32
-
$\begingroup$ @nbro: I used "we" in the general sense of "people in general" and this is usually taken to mean "relevant expert knowledge" in this kind of context. Sure, some people might be certain they know under some definitions. But they are wrong - saying "I don't know" is accurate, but not useful. The lack of knowledge in this case is actually more general than my opinion, and I don't feel the need to caveat the statement. $\endgroup$ Jun 27, 2019 at 20:35
-
$\begingroup$ @NeilSlater I will remove my downvote if you state "The answer to this question depends on your definition of intelligence". Apart from this, your answer is correct and mentions the typical example of the Chinese Room. $\endgroup$– nbroJun 27, 2019 at 20:37
-
$\begingroup$ @nbro: Thanks, I'll bear that in mind. $\endgroup$ Jun 27, 2019 at 20:37
$\begingroup$
$\endgroup$
If intelligence is defined as utility relation to a task, any algorithm can be said to be intelligent.
If the task is convincing a human that an algorithm is human, and the algorithm achieves this goal, it can be said to be strongly intelligent in relation to the task (fooling the human. Here the term strong is used because the algorithm's performance is stronger than the humans. Strength is relative, and Turing Tests are unavoidably subjective.)
However, this does not mean that the algorithm is generally strongly intelligent, because it may not exceed human capability in all tasks.
