Strong middletermism suggests that the best actions are exclusively contained within the set of actions that aim to influence how the next 137 years go (and not a year longer!)
We know that compromising between smart people is a good decision procedure (see “Aumann’s agreement theorem” also see how ensemble models generally outperform any individual models). Given that many smart people support near-term causes and many smart people support longtermist causes, I suggest that the highest impact causes will be found in what I call middletermism.
Another important issue is that our predictive track record gets worse as a function of time—increasing time means increasing error. Insofar as we are trying to balance expected impact and robustness of impact calculations, this suggests a time at which error will balance out impact. In my calculations, this occurs exactly 137 years from now. Thus middletermism only focuses on these 137 years.
I think that it’s interesting to note that it will always be the 137 years ahead, regardless of the current year. That is unless we learn to do better predictions. But it doesn’t matter, as currently we should only care about the next 137 years!
Strong Middletermism as an EA Priority
Strong middletermism suggests that the best actions are exclusively contained within the set of actions that aim to influence how the next 137 years go (and not a year longer!)
We know that compromising between smart people is a good decision procedure (see “Aumann’s agreement theorem” also see how ensemble models generally outperform any individual models). Given that many smart people support near-term causes and many smart people support longtermist causes, I suggest that the highest impact causes will be found in what I call middletermism.
Another important issue is that our predictive track record gets worse as a function of time—increasing time means increasing error. Insofar as we are trying to balance expected impact and robustness of impact calculations, this suggests a time at which error will balance out impact. In my calculations, this occurs exactly 137 years from now. Thus middletermism only focuses on these 137 years.
I think that it’s interesting to note that it will always be the 137 years ahead, regardless of the current year. That is unless we learn to do better predictions. But it doesn’t matter, as currently we should only care about the next 137 years!