Like the human brain, LLMs use separate sets of units for language, formal reasoning, social reasoning, and intuitive physical reasoning. A modular organization of cognition may be a fundamental principle of intelligence!
The human brain is strikingly modular: distinct networks for language, formal reasoning, social reasoning, physical reasoning. Is this fundamental to intelligent systems, or an accident of evolution? In our new preprint, we find the same modular organization emerges in LLMs.