Uh, the bellman equation was first used for control theory and is the foundation of modern reinforcement learning... so wouldn't that imply LLMs "come from" control theory?