Good post! Been thinking about distributed agents quite a bit, and what forms of coordination would be needed. Starting here, https://www.strangeloopcanon.com/p/seeing-like-an-agent but also more written with Alex Imas. We will need to a) learn quite a bit more about these agents themselves, and b) bring the understanding from economics and decision theory to figure out how to design institutions.
It's interesting how this awareness of the collective nature of intelligence is emerging as we start to be able to explore intelligence in a much more empirical way than when humans were the only example most people considered. Though my own intuition on this is a bit mixed since I think that collective intelligences produce centralized agents that behave more or less like one would expect them to. Those centralized agents are what Norbert Wiener called virtual governors; they embody preferences that organize the members of the system toward system-level goals. (https://interestingessays.substack.com/p/virtual-governors-and-arrows-impossibility) We'll have a paper coming out about this with applications to AI soon, hopefully!
I am interested to see a recording if there is one or if you plan to give this talk in the future.
Good post! Been thinking about distributed agents quite a bit, and what forms of coordination would be needed. Starting here, https://www.strangeloopcanon.com/p/seeing-like-an-agent but also more written with Alex Imas. We will need to a) learn quite a bit more about these agents themselves, and b) bring the understanding from economics and decision theory to figure out how to design institutions.
Lots of good points here, I hope the next talk gets recorded!
It's interesting how this awareness of the collective nature of intelligence is emerging as we start to be able to explore intelligence in a much more empirical way than when humans were the only example most people considered. Though my own intuition on this is a bit mixed since I think that collective intelligences produce centralized agents that behave more or less like one would expect them to. Those centralized agents are what Norbert Wiener called virtual governors; they embody preferences that organize the members of the system toward system-level goals. (https://interestingessays.substack.com/p/virtual-governors-and-arrows-impossibility) We'll have a paper coming out about this with applications to AI soon, hopefully!