I am a teacher of mathematics and physics, and I often have students who participate in competitions. For example, the first round of the Physics Olympiad in Germany is conducted as a homework assignment (the second round is an exam). Since the advent of ChatGPT, there’s been a problem: ChatGPT can solve such problems (or at least give the impression of doing so). The appeal of puzzling and competing as a homework format is lost when students are so easily tempted to ask an AI.
My question is: Is there a mechanism by which one can report certain problems to OpenAI? If someone then asks the AI these questions, ChatGPT would respond that the problem is from a current competition and therefore cannot provide a solution. I assume that the block could be extended to variations of the problem. If it’s possible to prevent ChatGPT from providing information on building dangerous weapons, there should be similar mechanisms for this issue.