Researchers conducted circuit analyses across 46 tasks spanning four cognitive domains to test whether large language models develop specialized internal architectures analogous to the brain’s functional networks. They found that tasks drawing on the same functional network in humans recruit overlapping neurons in LLMs, while tasks drawing on different networks recruit distinct neurons. The authors argue this convergent emergence of modularity across biological brains and artificial neural networks suggests functional specialization may be a fundamental organizing principle of intelligent systems generally, rather than a quirk specific to biological evolution.
