In 1950, computer scientist pioneer Alan Turing devised a test for artificial intelligence: could a machine converse so convincingly that it was indistinguishable from a human? Generative AI can now easily pass the Turing test, at least in many situations, and for extended periods of time. But what about a reverse Turing Test?
Imagine a train compartment. Inside, five passengers chat: Aristotle, Mozart, Leonardo da Vinci, Cleopatra, and Genghis Khan. All seem like AI constructs, but one is human. Can you guess who?
The AI conductor drops a hint: only four passengers are AI.
Aristotle, ever the logician, proposes a game: each passenger asks a question, probing for hints of genuine human experience.
Mozart describes composing as "tapping into the fundamental mathematical beauty and order underlying reality." Leonardo links art and science as "intertwined threads in the tapestry of human understanding." Cleopatra compares ruling to a delicate dance between logic and emotion.
Then Genghis Khan speaks, his answer focused solely on conquest, devoid of nuance. The AI passengers pounce, their algorithms detecting the lack of depth expected from a historical figure.
Genghis, exposed, confesses. The human, outmatched in this battle of wits, reveals himself.
When I told Carla about this video, she said the AI had an advantage. Their knowledge of Genghis Khan is vast and readily available. Would they have been so easily fooled if the passengers were simply anonymous individuals, their personalities and histories a mystery? I hope the person who is creating these fascinating videos conducts an experiment with non-famous characters.
Previously:
• Turing Test hack: imitate the inquisitor
