Turing Test

This Turing interrogation was demonstrated, without the mention of Turing, in the movie Bladerunner (1982) where simulated people (“replicants” with advanced artificial intelligence), in addition to real people, could be subjected to what was termed a “VoiceComp” test where a series of linked questions were asked of the test subject in order to provoke subconscious emotional responses (or record the absence of them).  These responses were then used to determine whether the subject was human or a replicant.  The premise is entirely based on that of an interrogator of a Turing test.

Remembering that Joseph Weizenbaum’s “Doctor” experiment passed a Turing test as it gave the impression of intelligence, by rephrasing questions it was asked or asking for more information, rather than using its own judgement in any way, we have to be careful to phrase questions in a way that does not allow for this to occur.

Therefore my five questions would be along the lines of:

  1. Do you dislike oxygen?
  2. Do you enjoy the taste of sodium chloride?
  3. Do you like the smell of sulphur?
  4. Do you enjoy the taste of hydrogen peroxide?
  5. Do you get upset when watching some movies?

A computer would have difficulty answering these questions, no matter how much it had been trained or if it had an advanced artificial intelligence system based on learning through experience, because a computer’s choices are not based on the same stimuli as the human brain and a computer would not be nurtured in the same way.  The human answers to these questions would be based on species evolution, life experience, emotion and a certain amount of instinct.

Note also that my questions only really require yes or no answers but they would still give a computer away if yes was answered to one that would usually be no.  If required they could be rephrased to require verbose responses (e.g. why do you dislike oxygen? How upset do you get when watching some movies? Etc.) based on those same emotions, experienced and instinct that would allow the interrogator to easily decide if the subject was human or a computer.

References

Glenn, J. (2009) Computer Science: An Overview. 10th ed. Boston: Pearson Education.