Right now, humans still have enough practice thinking to point out the errors, but what happens when humanity becomes increasingly dependent on LLMs to do this thinking?