I posted yesterday about how GPTZero was such a horrible idea, and now this nightmare. Detecting AI generated text is impossible without knowing what model was

I posted yesterday about how GPTZero was such a horrible idea, and now this nightmare. Detecting AI generated text is impossible without knowing what model was used. It could be more feasible for them to detect text written by their own models given that they know the logits. But OpenAI doesn’t have a monopoly on large language models. However, the consequences for false positives are so dire that I would never want to create such a tool. It will be misused, and hiding behind “it’s just information” is no excuse. You don’t admit testimony unless it’s reliable. At one time I thought that OpenAI was out to make the world a better place. But now it’s clear to me that ethics is the last thing on their mind.