gpt3.5-turbo fails the turing test due to it's constant butt covering. Davinci can pass for a human. I am speaking only of the API responses. The "chatgpt" web interface is something different.