Because it’s not fucking ai, it never was, and it turns out the actual real world use cases for systems that are extremely good at convincing you their statements are the truth regardless of if it is, aren’t that many.