# Funding unlocked might be a good way to measure the ROI on addressing bottlenecks [[ROI is tricky when you’re talking about advancing humanity]] in large part because it requires collapsing a lot of different and potentially illegible things into a single scalar value. However, if you scope the domain you care about to “unblocking chokepoints for progress” the ROI measurement “dollars unlocked per dollar spent” might actually do a good job capturing success, failure, and efficiency. When we’re talking about a “progress chokepoint” there’s an implicit corollary that “if work was done on this chokepoint, it would then unlock a lot more (useful) work in the same area.” Making that desired outcome explicit leads to the question “how could we measure unlocked work?” “Funding dollars spent that wouldn’t have been spent otherwise” makes sense as a way to measure unlocked work. Spent money is a reasonable measurement for work - people are paid for work and capital costs are like captured mechanized work. It actually captures “changed minds” as well because there’s little point of changing people’s minds if it doesn’t change actions eventually, so we would expect changed minds to eventually turn into funding from governments, etc. Concretely, a “unlocked dollars” metric would look something like “We have good reason to believe that funding Michael Nielsen could potentially lead to the creation of a new field that is as big a deal as quantum computing. Since its foundation in the 1980’s, quantum computing funding work has been funded to the order of $5B and by 2030 it will probably be funded to the order of $50B (I am making up these numbers), so we could roughly say that the first 50 years of a new important field is $50B (it’s probably important to have a time-scale on it). Thus, if the Nielsen project costs $100M, the potential ROI could be 500% (with error bars going down to zero and up to ?? If he starts multiple new fields.)” There are still a many flaws in the metric of course, but a lot of them are addressed by using “dollars unlocked” as only a forward-looking way to compare potential projects and *not* as a retrospective success metric.[[Counterfactuals are hard]], so you could always make the argument that follow-on money would have been spent anyways. Here, you need to trust the expertise of person doing the estimate that they’re pretty sure that without the intervention the follow-on money won’t be spent. As always this metric is subject to [[Goodhart's Law]] - if it becomes a retrospective measure of success than one could imagine spending a lot of effort to cajole people into follow-on funding a project that should be left to die. The solution is to avoid letting unlocked dollars become a retrospective success metric. Dollars unlocked wouldn’t capture an intervention that unlocked an army of volunteers - this seems like an acceptable hole. The metric will skew towards capital-intensive areas. If you unlock a lot of work to settle mars, much more money will be unlocked than if you unlock a new discipline of theoretical physics just because of the nature of the work. The metric will also rank most highly the things that are barely choked, which is not necessarily bad (the last bit of work to get over a barrier is as important as the first) but it will downgrade areas that need a lot of hard work to convince people of their validity, even if amazing things will happen afterwards. Despite its flaws, unlocked dollars per dollar has many advantages over something like “value created” (as a proxy for impact.) First, unlocked dollars only requires one or two causal leaps even for very research-y or unprofitable but important areas. By contrast, looking at value requires a lot of causal gymnastics to say what effects and what fractions of those effects are attributable directly to a person, organization, or project. Second, going for pure value pushes towards things with effectively infinite value like AGI - “if you create AGI, it will create everything else so you get to lump that into your value calculation.” Any finite fraction of infinity will still be infinity, so you’re forced to either surrender with the math or override the metric. Third, unlocked dollars doesn’t require you to convert non-monetary-but-good-things into dollars. Measuring pure value created requires you to ask ultimately philosophical questions like “what is the value of a year of the end of person’s life?” “What is the value of sailing a boat on Europa?” etc. [Web URL for this note](http://notes.benjaminreinhardt.com/Funding+unlocked+might+be+a+good+way+to+measure+the+ROI+on+addressing+bottlenecks) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Funding+unlocked+might+be+a+good+way+to+measure+the+ROI+on+addressing+bottlenecks)