Hi Nick, Great work getting so much interest and so many ideas.
I am super curious to know how much prioritisation and vetting is going on behind the scenes for the ideas on the FTX Fund project list and how confident you are in the specific ideas listed.
One way to express this would be: Doyou see the ideas on your list as likely to be in the top 100 longtermist project ideas or as likely to be in the top 10,000 longtermist project ideas or somewhere in between?* I think knowing this could be useful for anyone looking to start a project to decide how closely to stick to the ideas you list.
I expect part of the reason why a number of people commented that they were expecting more awards/winners (and I too had this intuition) is because it may not be at all clear to the causal EA reader why the ideas listed in the 723 comments replying to the competition post are any better or worse than the ideas listed on the FTX project list. I think it is possible that there is more vetting and prioritisation going on behind the scenes in FTX than people realise. But if so it would be good to make that transparent.
Thank you Nick!
– –
NOTE: For what it is worth, my initial intuition when reading though the list, is that all the FTX ideas were very likely in the top “top 2500 longtermist ideas”* across FTX’s areas of interest. Which is decent. My intuition was driven by focusing on topics where I have done ideas research**:
The FTX list included “Alternative voting systems”. Voting is the most outwardly obvious part of the the policy reflection and decision process, but is roughly just 1% of the process. So this would be a top 100 idea for improving policy reflection and decision process. (Within the broader category of “Values and Reflective Processes” maybe a top 300 idea).
The FTX list included “Strengthening the Bioweapons Convention”. This is perhaps in the top 50 for ideas preventing dangerous research based on this mapping, and a top 150 idea for preventing or mitigating an existential biorisks.
– –
* By saying an collection of ideas are “top 1000 ideas” I mean that I could with a bit of time write out or collect 1000 ideas in the relevant category(s) where it would not be obvious that they were better or worse than the ideas in question (but substantially more than 1000 would be difficult). So if I then spent some months doing prioritisation research there would be a roughly 10% chance that any specific idea from the collection would make it to the top 100 (the top 10%).
** E.g. through my current job which is running a team at Charity Entrepreneurship who work on listing, prioritising between and researching ideas for EA orgs.
I suspect a lot of the “very best” ideas in terms of which things are ex ante the best to do, if we don’t look at other things in the space (including things not currently done), will look very similar to each other.
Like 10 extremely similar AI alignment proposals.
So I’d expect any list to have a lot of regularization for uniqueness/side constraint optimizations, rather than thinking of the FTX project ideas list as a ranked list of the most important x-risk reducing projects on the margin. Arguably, the latter ought to be closer to how altruistic individuals should be optimizing for what projects to do, after adjusting for personal fit
Hi Nick, Great work getting so much interest and so many ideas.
I am super curious to know how much prioritisation and vetting is going on behind the scenes for the ideas on the FTX Fund project list and how confident you are in the specific ideas listed.
One way to express this would be: Do you see the ideas on your list as likely to be in the top 100 longtermist project ideas or as likely to be in the top 10,000 longtermist project ideas or somewhere in between?* I think knowing this could be useful for anyone looking to start a project to decide how closely to stick to the ideas you list.
I expect part of the reason why a number of people commented that they were expecting more awards/winners (and I too had this intuition) is because it may not be at all clear to the causal EA reader why the ideas listed in the 723 comments replying to the competition post are any better or worse than the ideas listed on the FTX project list. I think it is possible that there is more vetting and prioritisation going on behind the scenes in FTX than people realise. But if so it would be good to make that transparent.
Thank you Nick!
– –
NOTE: For what it is worth, my initial intuition when reading though the list, is that all the FTX ideas were very likely in the top “top 2500 longtermist ideas”* across FTX’s areas of interest. Which is decent. My intuition was driven by focusing on topics where I have done ideas research**:
The FTX list included “Alternative voting systems”. Voting is the most outwardly obvious part of the the policy reflection and decision process, but is roughly just 1% of the process. So this would be a top 100 idea for improving policy reflection and decision process. (Within the broader category of “Values and Reflective Processes” maybe a top 300 idea).
The FTX list included “Strengthening the Bioweapons Convention”. This is perhaps in the top 50 for ideas preventing dangerous research based on this mapping, and a top 150 idea for preventing or mitigating an existential biorisks.
– –
* By saying an collection of ideas are “top 1000 ideas” I mean that I could with a bit of time write out or collect 1000 ideas in the relevant category(s) where it would not be obvious that they were better or worse than the ideas in question (but substantially more than 1000 would be difficult). So if I then spent some months doing prioritisation research there would be a roughly 10% chance that any specific idea from the collection would make it to the top 100 (the top 10%).
** E.g. through my current job which is running a team at Charity Entrepreneurship who work on listing, prioritising between and researching ideas for EA orgs.
I suspect a lot of the “very best” ideas in terms of which things are ex ante the best to do, if we don’t look at other things in the space (including things not currently done), will look very similar to each other.
Like 10 extremely similar AI alignment proposals.
So I’d expect any list to have a lot of regularization for uniqueness/side constraint optimizations, rather than thinking of the FTX project ideas list as a ranked list of the most important x-risk reducing projects on the margin. Arguably, the latter ought to be closer to how altruistic individuals should be optimizing for what projects to do, after adjusting for personal fit