<ul><li>A research study conducted by the University of California, San Diego, reveals that OpenAI's GPT-4.5, an AI model, has passed a three-party Turing test with a win rate of 73% when prompted to adopt a human-like demeanor.</li><li>The study tested three AI models, including GPT-4.5, Meta's Llama 3.1 405B, and OpenAI's GPT-4o, in conversations with human participants using a split-screen interface.</li><li>Results showed that GPT-4.5 and Llama 3.1 405B with a human-like persona prompt achieved win rates of 73% and 56% respectively, while GPT-4o without a persona prompt achieved a win rate of 21%.</li><li>The researchers believe that AI systems like GPT-4.5 could replace or supplement human labor in tasks that require conversational exchanges.</li></ul>

GPT 4.5 Passes the Turing Test: Study

Discover more