Thoughts as probabilistic code

Cognitive scientist Evelina Fedorenko summarizes evidence that many human thoughts are best modeled not as inner speech or images but as compositions in a probabilistic 'language of thought' (PLOT) — a code‑like symbolic system instantiated in distinct brain circuits and testable with imaging, lesions, and computational models. She argues language is primarily a communication tool, not the substrate of general reasoning, and that studying LLMs can illuminate how non‑linguistic code‑like representations implement human problem solving. — If thought often uses a code‑like probabilistic format, debates about AI, education, mental health, and free‑speech policy should shift from analogies to natural language toward mechanisms that are computational and brain‑structured.