I've seen similar arguments before (for example here -> https://thoughtinfection.com/2014/04/19/capitalism-is-a-paperclip-maximizer/ https://thoughtinfection.co

I've seen similar arguments before (for example here -> https://thoughtinfection.com/2014/04/19/capitalism-is-a-paperclip-maximizer/ https://thoughtinfection.com/2014/04/19/capitalism-is-a-pape... ) and I think it's a good analogy. What I don't understand is how this gets presented as an argument against AI risk. If the "paperclip maximizers" we construct now out of humans, machines, institutional processes and legal codes already exist, are successful, exhibit runaway growth, are a danger and lack insight in Chiang's sense, isn't that actually evidence in favor of the viability of such entities? If they are already dangerous even with humans in the loop and relying on cumbersome institutional processes, isn't that all the more reason to think about how much more dangerous they can get as they become more autonomous and agile?