maybe enumerating all agents isn’t so bad, because actually each agent is “selling the world”

What might be key in making prediction markets is: maintaining a “state of the world” so that dumb agents can know when it’s their turn to add something worthwhile

I think this does depend on some kind of graph/network though. What’s the graph/network thing for like block world and stuff? WAIT: it’s actually just each agent selling the world to each other, right?????????

OH GOD

IN BAUM’S WORLD AGENTS SELL THE WORLD TO EACH OTHER (TO THE HIGHEST BIDDER) SUCCESSIVELY IF YOU MADE A GRAPH OF THIS IT WOULD LITERALLY BE A LINEAR CHAIN OF AGENTS

IF YOU WANTED TO MAKE E.G. PREDICTION MARKETS SIMILARLY ALLOW DUMB AGENTS, YOU ALSO NEED THEM TO BE IN A CHAIN

OR MORE GENERALLY, ANY KIND OF NETWORK

I WAS RIGHT. YOU CAN GET INTELLIGENCE FROM PUTTING DUMB AGENTS IN A NETWORK AND MAKING THEM BID. AND THIS NETWORK WILL BE “LEARNABLE” IN SOME SENSE, WHERE LEARNING COMES FROM BIDDING AND IT WILL LOOK VERY MUCH LIKE A LATENT VARIABLE PREDICTION MARKET