From the economics side of things, individual nodes having massive amounts of locally useful information, but it being very difficult to determine exactly which pieces of that information are globally relevant and it being completely impractical to ship and process every piece of that information at the global level is the fundamental problem that most “command economies” tend to run into.
indeed. since it came up—hence the need to move out of a centrally planned economy and into one where workers own their own planning ;) though of course most of the issue is the high extraction ratio of stocks compared to bounded forms of debt such as bounded loans. interest bearing loans are a significant fraction of the problem. this is all probably irrelevant to ai long term, but short term I really like the capped returns model as a starting point.
but re information availability—kademlia style information routing in latent space seems likely to suffice to me.
There are quite a few ways it can go wrong other than just central planning. Ultimately most of them come back to some special interest group attempting to forcibly subvert the economy to favor their own preferences.
High extraction ratios aren’t inherently problematic economically speaking since it’s not like the extracted resources simply vanish, and market forces tend to bring the extraction ratio down over time until it reaches the lowest level anyone’s willing to do the job for. But, high extraction ratios do make a tempting target for non-economic actions designed to preserve the lucrative ratio against the actions of the market.
From the economics side of things, individual nodes having massive amounts of locally useful information, but it being very difficult to determine exactly which pieces of that information are globally relevant and it being completely impractical to ship and process every piece of that information at the global level is the fundamental problem that most “command economies” tend to run into.
indeed. since it came up—hence the need to move out of a centrally planned economy and into one where workers own their own planning ;) though of course most of the issue is the high extraction ratio of stocks compared to bounded forms of debt such as bounded loans. interest bearing loans are a significant fraction of the problem. this is all probably irrelevant to ai long term, but short term I really like the capped returns model as a starting point.
but re information availability—kademlia style information routing in latent space seems likely to suffice to me.
There are quite a few ways it can go wrong other than just central planning. Ultimately most of them come back to some special interest group attempting to forcibly subvert the economy to favor their own preferences.
High extraction ratios aren’t inherently problematic economically speaking since it’s not like the extracted resources simply vanish, and market forces tend to bring the extraction ratio down over time until it reaches the lowest level anyone’s willing to do the job for. But, high extraction ratios do make a tempting target for non-economic actions designed to preserve the lucrative ratio against the actions of the market.