They imagine a simulation of themselves, but this simulation isn't necessarily realistic. They run the simulation, and the words the simulation uses to describe

They imagine a simulation of themselves, but this simulation isn't necessarily realistic. They run the simulation, and the words the simulation uses to describe itself "visualzing" they just repeat verbatim. Human consciousness and self-awareness are so dim that they mistake this for themselves being able to do the same. If someone didn't have this "skill", they could prompt an LLM to "visualize", then repeat the words off the screen, and but for the clues that they're cheating bystanders wouldn't be able to tell much difference. I assert that there is no additional insight gained by the "visualization" that isn't available from the verbalization because these are, in fact, essentially the same thing.