And additionally, whether our memory and long term learning - and even our goal-choosing - is fundamentally different from an indexed storage of strings of tokens that can be brought back into short-term context when “triggered” by their embedding-similarity to the current context.