That doesn’t really make sense. I think the linked argument is persuasive but it has little bearing on the situation at hand. (By the same token you could argue

That doesn’t really make sense. I think the linked argument is persuasive but it has little bearing on the situation at hand. (By the same token you could argue making that argument at all – thus providing this approach as a solution to problems to the long term memory of large language models – means you have least. This is getting into Roko’s Basilisk territory.) What you arguments amounts to is making it impossible to write critically about LLMs. Because writing “yeah, Bing did some weird shit but no, I can’t show you the transcript” just doesn’t fly in the real world. That won’t reach those who ultimately actually have power over it (companies and society at large).