Risk neutral grantmakers should, if they have not already, strongly consider modifying their position. If such a grantmaker has a choice of an intervention with 1000 utils of potential impact but only 1% chance of working out (10 utils in expectation), and an intervention with 10 utils of potential impact but 90% likely to work out (9 utils in expectation), I would suggest that one should go with the latter at this point where the x-risk community is hopefully still in its early days.
The reason is that having wins has value in and of itself. I think this is especially true in the x-risk domain where the path to impact is uncertain and complex. At least now, in the hopefully early days of such work, there might be significant value by just demonstrating to ourselves, and perhaps major donors on the fence on whether to become “EA/x-risk donors” and also perhaps talent wondering if EA “is for real”, that we can do something.
You are not going to be able to get an accurate estimate for the “value of a marginal win”.
I also doubt that you can accurately estimate a “1% chance of 1000 utilitons”. In my opinion guesses like these tend to be based on flimsy assumptions and usually wildly overestimated.
What do you mean by “accurate estimate”? The more sophisticated version would be to create a probability distribution over the value of the marginal win, as well as for the intervention, and then perform a Monte-Carlo analysis, possibly with a sensitivity analysis.
But I imagine your disagreement goes deeper than that?
In general, I agree with the just estimate everything approach, but I imagine you have some arguments here.
Risk neutral grantmakers should, if they have not already, strongly consider modifying their position. If such a grantmaker has a choice of an intervention with 1000 utils of potential impact but only 1% chance of working out (10 utils in expectation), and an intervention with 10 utils of potential impact but 90% likely to work out (9 utils in expectation), I would suggest that one should go with the latter at this point where the x-risk community is hopefully still in its early days.
The reason is that having wins has value in and of itself. I think this is especially true in the x-risk domain where the path to impact is uncertain and complex. At least now, in the hopefully early days of such work, there might be significant value by just demonstrating to ourselves, and perhaps major donors on the fence on whether to become “EA/x-risk donors” and also perhaps talent wondering if EA “is for real”, that we can do something.
Isn’t the solution to this to quantify the value of a marginal win, and add it to the expected utility of the intervention?
You are not going to be able to get an accurate estimate for the “value of a marginal win”.
I also doubt that you can accurately estimate a “1% chance of 1000 utilitons”. In my opinion guesses like these tend to be based on flimsy assumptions and usually wildly overestimated.
What do you mean by “accurate estimate”? The more sophisticated version would be to create a probability distribution over the value of the marginal win, as well as for the intervention, and then perform a Monte-Carlo analysis, possibly with a sensitivity analysis.
But I imagine your disagreement goes deeper than that?
In general, I agree with the just estimate everything approach, but I imagine you have some arguments here.