How much worse is 1.0 than 1.1 in practical situations?
This isn’t a very salient question in my mind, because at this point I think it’s better to think of both UDT 1.0 and 1.1 as introducing new problems, as opposed to introducing solutions or even foundations for future solutions to build on (i.e., the real solution might look very different). Basically UDT 1.0 says we seemingly have to be updateless and also take into account logical correlations between different agents or instances of an agent, and UDT 1.1 says that doesn’t seem enough to even coordinate copies of one agent; the one-agent problem can be solved through optimizing global strategy but it’s totally unclear how to handle multiple agents. See this quote from the UDT 1.1 post:
What about when there are agents with difference source codes and different preferences? The result here suggests that one of our big unsolved problems, that of generally deriving a “good and fair” global outcome from agents optimizing their own preferences while taking logical correlations into consideration, may be unsolvable, since consideration of logical correlations does not seem powerful enough to always obtain a “good and fair” global outcome even in the single-player case. Perhaps we need to take an approach more like cousin_it’s, and try to solve the cooperation problem from the top down. That is, by explicitly specifying a fair way to merge preferences, and simultaneously figuring out how to get agents to join into such a cooperation.
This isn’t a very salient question in my mind, because at this point I think it’s better to think of both UDT 1.0 and 1.1 as introducing new problems, as opposed to introducing solutions or even foundations for future solutions to build on (i.e., the real solution might look very different). Basically UDT 1.0 says we seemingly have to be updateless and also take into account logical correlations between different agents or instances of an agent, and UDT 1.1 says that doesn’t seem enough to even coordinate copies of one agent; the one-agent problem can be solved through optimizing global strategy but it’s totally unclear how to handle multiple agents. See this quote from the UDT 1.1 post: