They don’t really demonstrate general intelligence. They’re very powerful tools, but LLMs are still a form of specialized intelligence, they’re just specialized at language instead of some other task. I do agree that they’re closer than what we’ve seen in the past, but the fact that they don’t actually understand our world and can only mimic the way we talk about it still occasionally shines through.
You wouldn’t consider Midjourney or Stable Diffusion to have general intelligence because they can generate accurate pictures of a wide variety of things, and in my opinion, LLMs aren’t much different.