My UDT 1.1 is a predecessor to FDT and similar to it in many ways, but did not use counterfactuals/counterpossibles, instead relying on epistemic conditioning (of logical uncertainty) to calculate expected utilities. You can think of it as a version of updateless EDT (which you seem to favor) that nails down some details of how it might work.
I do think some people are perhaps too confident or eager about “adopting FDT”. In my view it’s not even alpha stage in analogy with software development. As I wrote in a previous post, it seems better to frame UDT/FDT as introducing some problems that weren’t previously considered or paid much attention to (e.g., need for / how to do updatelessness, issues with indexical selfishness, what DT is good for AIs, see here and here for more), along with our current attempts to solve them.
How much worse is 1.0 than 1.1 in practical situations? UDT 1.1 has excellent theoretical properties, but to my knowledge hasn’t been useful for further work so far. Non exhaustively, this seems to be because attempts to reduce the size of the outer loop and spread the computational work out over time don’t work properly in 1.1, due to the use of an optimal global strategy.
My UDT 1.1 is a predecessor to FDT and similar to it in many ways, but did not use counterfactuals/counterpossibles, instead relying on epistemic conditioning (of logical uncertainty) to calculate expected utilities. You can think of it as a version of updateless EDT (which you seem to favor) that nails down some details of how it might work.
See my reply to Will MacAskill about his Bomb thought experiment.
I do think some people are perhaps too confident or eager about “adopting FDT”. In my view it’s not even alpha stage in analogy with software development. As I wrote in a previous post, it seems better to frame UDT/FDT as introducing some problems that weren’t previously considered or paid much attention to (e.g., need for / how to do updatelessness, issues with indexical selfishness, what DT is good for AIs, see here and here for more), along with our current attempts to solve them.
How much worse is 1.0 than 1.1 in practical situations? UDT 1.1 has excellent theoretical properties, but to my knowledge hasn’t been useful for further work so far. Non exhaustively, this seems to be because attempts to reduce the size of the outer loop and spread the computational work out over time don’t work properly in 1.1, due to the use of an optimal global strategy.