Great post, I wish more longtermist grantmakers wrote about this.
To copy over some of my comments:
My general reaction is that while I agree that prioritization is more important between buckets (especially the “future-improvement” conversion) than within them, the BOTEC maximalism framework as illustrated with these examples is much less convincing between buckets than within, which seems like a mismatch. Founders Pledge’s climate grantmaking team elides this entirely while remaining quantitative by exploiting shared structure in uncertainties across interventions to make robust-ish judgments about relative comparisons without necessarily committing to any unit conversions and such, seems worth contrasting. (You did note that you do this too when cheap.)
You also do mention having a bunch of cached takes on how good various intermediate desiderata are and how those all cash out to future-improvement, but given that you can’t publish them I have to take your word for it, which I guess is fine. I would be keen to see others publish it though, since Nuno Sempere’s experiment to elicit valuations of research from Fin Moorhouse, Gavin Leech, Jaime Sevilla, Linch Zhang, Misha Yagudin, and Ozzie Gooen using a pairwise comparison-based utility function extractor app led to both inconsistent preferences and wide-ranging disagreements, which makes me think people should compare notes on this more:
I also find it kind of wild that you can find many decent opportunities in the 1-20x range on your bar since I’d anchored to Linch’s old $100M per bp (0.2-0.9x) being a bar at which he’d feel bullish about funding, which I guess means I still haven’t internalised how wide the the between-buckets range can get, or it could mean you’re more like Gavin above (widest valuation range) than Fin (narrowest), and I don’t really know where to land on this valuation range spectrum.
Great post, I wish more longtermist grantmakers wrote about this.
To copy over some of my comments:
My general reaction is that while I agree that prioritization is more important between buckets (especially the “future-improvement” conversion) than within them, the BOTEC maximalism framework as illustrated with these examples is much less convincing between buckets than within, which seems like a mismatch. Founders Pledge’s climate grantmaking team elides this entirely while remaining quantitative by exploiting shared structure in uncertainties across interventions to make robust-ish judgments about relative comparisons without necessarily committing to any unit conversions and such, seems worth contrasting. (You did note that you do this too when cheap.)
You also do mention having a bunch of cached takes on how good various intermediate desiderata are and how those all cash out to future-improvement, but given that you can’t publish them I have to take your word for it, which I guess is fine. I would be keen to see others publish it though, since Nuno Sempere’s experiment to elicit valuations of research from Fin Moorhouse, Gavin Leech, Jaime Sevilla, Linch Zhang, Misha Yagudin, and Ozzie Gooen using a pairwise comparison-based utility function extractor app led to both inconsistent preferences and wide-ranging disagreements, which makes me think people should compare notes on this more:
I also find it kind of wild that you can find many decent opportunities in the 1-20x range on your bar since I’d anchored to Linch’s old $100M per bp (0.2-0.9x) being a bar at which he’d feel bullish about funding, which I guess means I still haven’t internalised how wide the the between-buckets range can get, or it could mean you’re more like Gavin above (widest valuation range) than Fin (narrowest), and I don’t really know where to land on this valuation range spectrum.