By testing humans against a broad range of models (including Gemini, GPT, and Llama) on standard creativity tasks—such as finding alternative uses for common objects—the researchers found that AI ...