AI agents engage in a unique challenge to uncover a human infiltrator—here’s how they do it Five AI models, embodying historical figures like Aristotle and Cleopatra, are aboard a train compartment with one concealed human. This setup forms the backdrop for a viral video where AI competes in a “reverse Turing test” against a human. Though AI emerges victorious, what insights does this provide into the realms of human and artificial intelligence? The traditional Turing test, proposed by Alan Turing, evaluates a machine’s ability to exhibit intelligence indistinguishable from a human’s. While no AI has definitively passed, recent claims suggest GPT-4’s success in a preprint study. In this reversed scenario, AI chatbots—portraying Aristotle, Mozart, da Vinci, Cleopatra, and Genghis Khan—interrogate one another. The human, impersonating Genghis Khan, delivers a quote from Conan the Barbarian, revealing their identity. AI detects the lack of nuance, voting against the human. To orchestrate the test, VR developer Tore Knabe scripted the dialogue and provided AI with conversational context. Despite the outcome favoring AI, experts like Anders Sandberg from Oxford caution that staged elements may influence perceptions. “Turing’s proposal provokes nuanced considerations on intelligence perception,” says Huma Shah from Coventry University, highlighting the multifaceted nature of assessing machine capabilities. As AI advancements continue, the quest to define and measure intelligence persists across diverse applications—from caregiving robots to autonomous vehicles.
