I don't think it is. The distribution of an AI model that was trained on such a huge amount of movie reviews is very close to the human distribution.
At least that's true around the mean. If your application needs to handle long-tail cases, an LLM won't easily give you that. But depending on the application, that may not be necessary. So yeah, sometimes this is a bad idea, but for many applications it may be just fine.
I don't think it is. The distribution of an AI model that was trained on such a huge amount of movie reviews is very close to the human distribution.
At least that's true around the mean. If your application needs to handle long-tail cases, an LLM won't easily give you that. But depending on the application, that may not be necessary. So yeah, sometimes this is a bad idea, but for many applications it may be just fine.