I find that the people most impressed by large language models are the liberals/leftists because they can be perfectly simulated by a chat bot. Both are talking complete nonsense while being utterly convinced that they are right.
For example I recently asked ChatGPT to give me a mathematical function that includes the gamma function and evaluates to 24.4. Not only where the answers not even close, at one point it claimed that 24.4 is not a positive real number. Then it gave me a python snippet to prove its function's correctness, but the program produced a value of about 4.9. ChatGPT doesn't even evaluate its own output.
The difference between a chat bot and a liberal is that the bot will always apologize profusely when I point out an error.