The "Turing Test", as we talk about today, is a very simplified take on what Turing described, probably in an effort to pass it. You can read his original paper here [1]. In the original test, a person of some role or identity or whatever would be introduced to the interrogator. It would then be up to the AI to imitate this identity, and the interrogator would have to pick the real person vs the impersonator.<p>The modern version of "human or AI" is rather dumbed down because all it requires is a passable chatbot. The example Turing offered was that of a woman. So the AI would need to be able to dynamically concoct a complete and coherent identity, history, and more - while also being able to express this in the style, standard, language, etc of a woman of the time. Also, perhaps just as relevantly, the bot would need to know and be able to mimic what the woman would <i>not</i> know given her backstory. Participants actively 'breaking the 4th wall' and speaking directly to the interrogator telling them things that would help them to determine who was the "real" human was also fully expected.<p>There's also an ironic human context to current Turing tests. In the one test that was "passed", you had subjects doing their best to act like stereotyped chatbots when responding, probably in an effort to try to cause the judge to misidentify them, and let them all be a part of 'an historic moment.' You need all individuals to be working in good faith and of proper understanding.<p>Would this test actually prove "thinking"? Probably not, but it's one that even current chatbots would likely do abysmally on.<p>[1] - <a href="https://redirect.cs.umbc.edu/courses/471/papers/turing.pdf" rel="nofollow">https://redirect.cs.umbc.edu/courses/471/papers/turing.pdf</a>