>you've mentioned an ongoing human fear, machines deciding what you think.

There are all kinds of machines that tell us what to think. I would consider any system that abstracts away the human to be a machine in this case. Society itself is one of these machines.

Language is possibly one of the most important things people can have a working knowledge of, especially now that there is so much of it. When you send a prompt to an LLM you're telling it what to think. When it sends text back, its telling you what to think, but you're at a disadvantage, when you think it changes you. The LLM outside of its context is read only.