I was working with Google Gemini on something, and I realized the AI talks to itself often because that's the only way it can remember its "thoughts". I was wondering why you don't have an AI write to an invisible "thoughts" box to think through a problem, and then write to the user from its thoughts? This could be used to do things such as emulate human thinking in chat bots, where it can have a human thought process invisibly, and write the results of the human-like thinking to the user.
Sorry if this is stupid, I'm a programmer and not incredibly experienced in AI networks.
[link] [comments]