One of the takeaways from Sejnowski’s paper [Sej] is that the starting prompts the AI is fed have a great deal to do with how it responds to certain topics. For example, telling the AI “you are a research scientist who is interested in accuracy and cited references and will not simply make up an answer if you do not know it.” See the posting [stderr]
Try playing 20 questions and give us a transcript, that should be good
Hmm. Is asking chatgpt whether a response is pre-programmed a bit like asking it if its citations are real or made up?
One of the takeaways from Sejnowski’s paper [Sej] is that the starting prompts the AI is fed have a great deal to do with how it responds to certain topics. For example, telling the AI “you are a research scientist who is interested in accuracy and cited references and will not simply make up an answer if you do not know it.” See the posting [stderr]