The whole vibe with LLMs is “wow, this is much better at language than anything else, surely it must be great for something?”
The problem is that “this will create much more value than it costs or destroys” is a completely unproven hypothesis and early data is not promising.