When artificial systems produce human-like language, people may draw a reverse inference: if LLMs can speak like humans, perhaps humans think like LLMs.
I think I experienced this when I learned about LLMs, chain of thought, thinking tokens, short-term memory context, and long-term memory context. I began applying these concepts to real life and reasoning about how our brains work as if these concepts described how our brains actually function. But maybe this is more akin to the Tetris effect?
People have been doing this since the invention of clockwork. Analogies are useful, even when they're utterly wrong, since they provide a perspective and that perspective is not necessarily wrong. Who knew?
Comments
I think I experienced this when I learned about LLMs, chain of thought, thinking tokens, short-term memory context, and long-term memory context. I began applying these concepts to real life and reasoning about how our brains work as if these concepts described how our brains actually function. But maybe this is more akin to the Tetris effect?
People have been doing this since the invention of clockwork. Analogies are useful, even when they're utterly wrong, since they provide a perspective and that perspective is not necessarily wrong. Who knew?