Thanks for writing this up. I might suggest targeting much more precisely than âstartupsâ/ââentrepreneursâ.
You say:
Anecdotally, many entrepreneurs bounce off the EA community, which has been described as overly risk-averse, too academic and theoretical, and âbiased against actionâ.
My personal experience with this complaint is that itâs usually a feature of reality, not the EA community. Even the most applied AI safety research agendas involve a lot more thinking about linear algebra than they do building a sales team.
Taking Future Fundâs list of projects as an example: most of them seem to require very substantial technical domain expertise. E.g. creating better PPE is probably the sort of project which would benefit from an entrepreneur who has started a medical device manufacturing business, but might be even more bound on scientists who can research different materials etc. (and is very much not bound on e.g. entrepreneurs who know a bunch about B2B marketing).
Some of this comment comes from my personal experience: itâs pretty rare that I feel like my impact could be increased through being better at the sorts of things which made me better at running a tech startup, and much more common that there is a specific technical skill set which I lack.
tl;dr: most âentrepreneursâ arenât a good fit for starting the highest impact projects, and a lot of what EAs describe as âstartupsâ are wildly different from sort of thing VCs fund.
I think this is an important point and thereâs some truth to it: EA âstartupsâ are different from venture-backed startups. When you say âtargetingâ, what do you mean here exactly, and do you have any ideas on how to do this without increasing time costs a lot? An advantage of our proposed approach is that itâs time-cheap because itâs systematic (with some caveats, we offer investment if and only if someone else has). But if there were a cheap way to make this better at finding the people most helpful for EA projects, that would be great.
---
If the claim is that, as a result of the differences between EA and venture-backed âstartupsâ, itâs not worth doing something like this to attract entrepreneurs to EA, Iâm not sure I agree:
1. There are some common skills to building both types of startupâyouâre creating a new org to do something untested with a small number of people and often not much money, so there are at least some similarities. Maybe the overlapping skillset is quite small. Do you think you had already built that skillset before leaving your startup, such that any extra skill-building in that overlapping core wouldnât be helpful for your work now? Iâm unsure about this because I havenât really built a startup of either type.
2. Itâs not just reality but also the EA community. a. Not everything can or should be tested by building a minimum viable product and seeing what reality throws at it, but for some types of problems this is useful. And, even for these types of problems, I think the default approach most EAs have is usually that problems are best solved by doing research and analytical thought instead. I think this probably leads to a bias against trying things. (We think hard when itâs appropriate, but also often when it would be better to just try it and see.) b. I know of at least one example of bias against action in the communityâsomeone applied for funding to distribute EA books at UPenn and said it was initially declined because CEA was concerned about possible downside risks. Both they and I thought this seemed crazy. c. Technical skills are required for many of the types of startups that VCs fund, but there are few technically-skilled entrepreneurs in the EA community. This suggests EA is pushing away even entrepreneurs who could contribute to technically demanding projects.
3. Earning to give can still be high impact, especially if youâre right and a traditional entrepreneur wouldnât be well-placed to start an EA (mega)project. If it works, Snowball Fund could either beat market returns (increasing the money in EA) or pay a small price (difference between market returns and its returns) per entrepreneur. If this means a higher number of EA-aligned entrepreneurs earning-to-give, or more giving from each, it may still be very cost-effective.
Hmmm, reading your question, I think our disagreement might not actually be about entrepreneursâ fit in EA, but rather about how much labor your plan requires. I think (correct me if Iâm wrong) you and I both believe something like:
If Snowball would require large amounts of labor to get close to market returns or to get impactful people involved in EA, then it might not be worth doing.
Let me briefly explain why I think the antecedent holds:
Market returns: Many (maybe most) VCs follow your strategy of being a follow-on investor, for the reasons you describe. Despite this, they still have to do a fair amount of labor and struggle to make market returns (e.g. because the most lucrative deals are only available to top VC firms).
Getting people involved: EAâs have a number of connections to talented entrepreneurs: several EA organizations have gone through YCombinator, many impressive entrepreneurs have signed with Founders Pledge, etc. Iâm struggling to think of instances where these networks resulted in recruiting someone to start an EA project. No doubt more could be done, and perhaps your team is more talented than the people who have tried this before, but I still doubt it will be easy.
Do you think you had already built that skillset before leaving your startup, such that any extra skill-building in that overlapping core wouldnât be helpful for your work now?
To be clear: itâs good to be better at more things, and of course I would be more impactful if I was better at start up stuff. But it just doesnât seem like the biggest thing holding me, or the people I know, back.
On point #1, thereâs a critical distinction in the type of âfollow-onâ strategy weâre employing, which is a standard template $100K into a large number of companies (generally known as the âspray and prayâ model). This is characterized by low diligence per deal as opposed to most VCâs who do still put in a decent amount of effort. Spray and pray of course has itâs drawbacks in terms of validating the quality of the deal flow, but thatâs the crux weâre exploring here regarding EA-aligned founders being potentially higher quality than average.
On point #2 not to be autobiographical but I personally am an example of an individual who would not have started an EA project without the onramp from Founders Pledge to actually introduce me to the community networks to get Snowball Fund up and running.
Thanks for writing this up. I might suggest targeting much more precisely than âstartupsâ/ââentrepreneursâ.
You say:
My personal experience with this complaint is that itâs usually a feature of reality, not the EA community. Even the most applied AI safety research agendas involve a lot more thinking about linear algebra than they do building a sales team.
Taking Future Fundâs list of projects as an example: most of them seem to require very substantial technical domain expertise. E.g. creating better PPE is probably the sort of project which would benefit from an entrepreneur who has started a medical device manufacturing business, but might be even more bound on scientists who can research different materials etc. (and is very much not bound on e.g. entrepreneurs who know a bunch about B2B marketing).
Some of this comment comes from my personal experience: itâs pretty rare that I feel like my impact could be increased through being better at the sorts of things which made me better at running a tech startup, and much more common that there is a specific technical skill set which I lack.
tl;dr: most âentrepreneursâ arenât a good fit for starting the highest impact projects, and a lot of what EAs describe as âstartupsâ are wildly different from sort of thing VCs fund.
I think this is an important point and thereâs some truth to it: EA âstartupsâ are different from venture-backed startups. When you say âtargetingâ, what do you mean here exactly, and do you have any ideas on how to do this without increasing time costs a lot? An advantage of our proposed approach is that itâs time-cheap because itâs systematic (with some caveats, we offer investment if and only if someone else has). But if there were a cheap way to make this better at finding the people most helpful for EA projects, that would be great.
---
If the claim is that, as a result of the differences between EA and venture-backed âstartupsâ, itâs not worth doing something like this to attract entrepreneurs to EA, Iâm not sure I agree:
1. There are some common skills to building both types of startupâyouâre creating a new org to do something untested with a small number of people and often not much money, so there are at least some similarities. Maybe the overlapping skillset is quite small. Do you think you had already built that skillset before leaving your startup, such that any extra skill-building in that overlapping core wouldnât be helpful for your work now? Iâm unsure about this because I havenât really built a startup of either type.
2. Itâs not just reality but also the EA community.
a. Not everything can or should be tested by building a minimum viable product and seeing what reality throws at it, but for some types of problems this is useful. And, even for these types of problems, I think the default approach most EAs have is usually that problems are best solved by doing research and analytical thought instead. I think this probably leads to a bias against trying things. (We think hard when itâs appropriate, but also often when it would be better to just try it and see.)
b. I know of at least one example of bias against action in the communityâsomeone applied for funding to distribute EA books at UPenn and said it was initially declined because CEA was concerned about possible downside risks. Both they and I thought this seemed crazy.
c. Technical skills are required for many of the types of startups that VCs fund, but there are few technically-skilled entrepreneurs in the EA community. This suggests EA is pushing away even entrepreneurs who could contribute to technically demanding projects.
3. Earning to give can still be high impact, especially if youâre right and a traditional entrepreneur wouldnât be well-placed to start an EA (mega)project. If it works, Snowball Fund could either beat market returns (increasing the money in EA) or pay a small price (difference between market returns and its returns) per entrepreneur. If this means a higher number of EA-aligned entrepreneurs earning-to-give, or more giving from each, it may still be very cost-effective.
Hmmm, reading your question, I think our disagreement might not actually be about entrepreneursâ fit in EA, but rather about how much labor your plan requires. I think (correct me if Iâm wrong) you and I both believe something like:
Let me briefly explain why I think the antecedent holds:
Market returns: Many (maybe most) VCs follow your strategy of being a follow-on investor, for the reasons you describe. Despite this, they still have to do a fair amount of labor and struggle to make market returns (e.g. because the most lucrative deals are only available to top VC firms).
Getting people involved: EAâs have a number of connections to talented entrepreneurs: several EA organizations have gone through YCombinator, many impressive entrepreneurs have signed with Founders Pledge, etc. Iâm struggling to think of instances where these networks resulted in recruiting someone to start an EA project. No doubt more could be done, and perhaps your team is more talented than the people who have tried this before, but I still doubt it will be easy.
To be clear: itâs good to be better at more things, and of course I would be more impactful if I was better at start up stuff. But it just doesnât seem like the biggest thing holding me, or the people I know, back.
Danielâs comments:
On point #1, thereâs a critical distinction in the type of âfollow-onâ strategy weâre employing, which is a standard template $100K into a large number of companies (generally known as the âspray and prayâ model). This is characterized by low diligence per deal as opposed to most VCâs who do still put in a decent amount of effort. Spray and pray of course has itâs drawbacks in terms of validating the quality of the deal flow, but thatâs the crux weâre exploring here regarding EA-aligned founders being potentially higher quality than average.
On point #2 not to be autobiographical but I personally am an example of an individual who would not have started an EA project without the onramp from Founders Pledge to actually introduce me to the community networks to get Snowball Fund up and running.
Thanks for the data point about you coming from FP! Thatâs helpful.