Can AI Evolve to Have More Natural Conversations?

I'm curious if there's a future for AI where it can handle more complex conversational dynamics than just basic input and output. For example, could an AI recognize multiple faces and voices at the same time? What if it could hear when several people are talking at once and express confusion if needed instead of sticking to a single response? Additionally, could it pause its speech when it realizes someone isn't done talking, and adapt to that situation? Finally, can AI start talking on its own based on reasoning about its environment, rather than just responding to prompts? Is there a current path or technology leading us toward these capabilities?