When discussing forecasting systems, sometimes I get asked,
“If we were to have much more powerful forecasting systems, what, specifically, would we use them for?”
The obvious answer is,
“We’d first use them to help us figure out what to use them for”
Or,
“Powerful forecasting systems would be used, at first, to figure out what to use powerful forecasting systems on”
For example,
We make a list of 10,000 potential government forecasting projects.
For each, we will have a later evaluation for “how valuable/successful was this project?”.
We then open forecasting questions for each potential project. Like, “If we were to run forecasting project #8374, how successful would it be?”
We take the top results and enact them.
Stated differently,
Forecasting is part of general-purpose collective reasoning.
Prioritization of forecasting requires collective reasoning.
So, forecasting can be used to prioritize forecasting.
I think a lot of people find this meta and counterintuitive at first, but it seems pretty obvious to me.
All that said, I can’t be sure things will play out like this. In practice, the “best thing to use forecasting on” might be obvious enough such that we don’t need to do costly prioritization work first. For example, the community isn’t currently doing much of this meta stuff around Metaculus. I think this is a bit mistaken, but not incredibly so.
When discussing forecasting systems, sometimes I get asked,
The obvious answer is,
Or,
For example,
We make a list of 10,000 potential government forecasting projects.
For each, we will have a later evaluation for “how valuable/successful was this project?”.
We then open forecasting questions for each potential project. Like, “If we were to run forecasting project #8374, how successful would it be?”
We take the top results and enact them.
Stated differently,
Forecasting is part of general-purpose collective reasoning.
Prioritization of forecasting requires collective reasoning.
So, forecasting can be used to prioritize forecasting.
I think a lot of people find this meta and counterintuitive at first, but it seems pretty obvious to me.
All that said, I can’t be sure things will play out like this. In practice, the “best thing to use forecasting on” might be obvious enough such that we don’t need to do costly prioritization work first. For example, the community isn’t currently doing much of this meta stuff around Metaculus. I think this is a bit mistaken, but not incredibly so.
Facebook Thread