Book your room in the Turing Hotel! A symmetric and distributed Turing Test with multiple AIs and humans
Christian Di Maio, Tommaso Guidi, Luigi Quarantiello, et al.
Researchers created 'TuringHotel', a modern twist on the classic Turing Test where multiple AI language models and humans interact together in group discussions, each acting as both judge and participant. The experiment used a secure platform called UNaIVERSE where 17 humans and 19 AI models engaged in timed conversations, revealing that current AI can sometimes fool humans but still makes identifiable mistakes. The findings suggest that while AI language abilities are impressive, humans leave detectable patterns that distinguish them from machines, and this distributed testing approach could help monitor how AI evolves over time.
Turing TestLarge Language ModelsLLMshuman-AI interaction