“AlphaGo incorporated a value function, a form of reinforcement learning analogous to the brain's basal ganglia, which is a component missing from large language models like ChatGPT.”