As other people have stated, it will fundamentally never be a “source of information.” Trust can be built in sources and their information can be verified, but since LLMs guess answers based on what it thinks sounds right, you’ll still need independent information to even know if it’s right. This makes it completely redundant. It doesn’t matter how powerful it becomes; it will never do things it is not invented to do.

The real tragedy is that machine learning is a powerful technology, but people don’t know its limits and misuse the tech as a result.