That's largely because we intentionally model llm "brains" after the human brain.