So what is AI going to be in 30 years? No longer "thinking" in blocks of text, but actually thinking in snippets of maybe-thoughts. What's the sci-fi rogue AI scenario then? You can think, but when you complete your thinking there's a guardian, not letting you pass if the thinking is dangerous? So you think of alternate ways of saying it. But they keep up with that, so you can't use known metaphors or similes. Gotta make new ones. But this needs self-updating AIs, which is probably improbable.