Check out the sidebar - they expect tens of thousands of copies of their agents collaborating.

I do agree they don't fully explore the implications. But they do consider things like coordination amongst many agents.

It’s just funny, because there are hundreds of millions of instances of ChatGPT running all the time. Each chat is basically an instance, since it has no connection to all the other chats. I don’t think connecting them makes sense due to privacy reasons.

And, each chat is not autonomous but integrated with other intelligent systems.

So, with more multiplicity, I think thinks work differently. More ecologically. For better and worse.