GPT-4o AI Agents Simulate 1,052 People's Behavior with 85% Accuracy

Nov 20, 2024, 07:01 PM

Researchers have utilized the latest version of OpenAI's ChatGPT, known as GPT-4o, to conduct two-hour interviews with 1,052 people. The AI agents, after receiving the transcripts, were tasked with simulating the interviewees' behavior and decisions, achieving an impressive 85% accuracy. This experiment, reported by New Scientist, showcases the potential of AI in replicating human behavior, offering an alternative to traditional focus groups and polling methods.