Hey there, I’m Austin, currently running https://manifund.org. Always happy to meet people; reach out at akrolsmir@gmail.com!
Austin
Karma: 4,748
Podcast: Austin and Oli on funding & incubating projects
Surplus: for massive public good
Sixteen schemes for AI safety
I’m unfortunately not interested in providing feedback on primarily AI-generated work or replying to AI-generated comments (Pangram flags the entirety of your response as AI-generated).
If you have specific questions or critiques you will write in your own words, I’m open to reading & engaging.
Hey! In response to what we judged to be low quality LLM submissions, we changed the Airtable view to only feature submissions we thought were past a particular quality bar (not strictly a filter on LLM usage, but there was definitely some correlation.)
Yours was not the only essay we filtered from the Airtable view, I would estimate about half were filtered.
I continue to be optimistic about LLM assistance for producing good essays, but I’d say the base rates are not great unfortunately.
I agree disclosure is for the benefit of the reader—I’m saying that, as a reader, I disprefer having to skip through a sentence at the top of many new posts disclaming that they used LLMs for copy editing and feedback.
I think the main thing I care about is “were large sections of this written directly by LLM” which I would prefer as first sentence so I know when to not read (which is actually the policy as written here, though I only realized that as of writing this comment). But—it appears that the default warning box has started scaring people into disclosing all forms of LLM usage at the top of essays, which I argue is a bad norm.
Disclosure is a reasonable idea, but mandating it at the top is awful, because the first line of a essay generally should be a hook, or convey the most information about the essay (after the title, anyways; especially because EA Forum doesn’t have a subtitle the way eg Substack does).
I would recommend allowing the author to put the disclosure anywhere in their essay. After the intro section might be a more natural place, or at the bottom similar to acknowledgements.
I don’t think there’s a settled consensus on the question of “should individual EAs split donations”; see Jeff Kaufman here and Eric Neyman/JP Addison here.
I used to be more on the side of “math out the impact EV and just give everything to one charity each year” but I’m now much closer to Nick’s recommendation of splitting your donations. Some arguments in favor of splitting:
From an ecosystem perspective, it’s better for a charity to have 100 $1k donors than to rely on a single $100k donor, both in terms of resilience and in terms of public transparency
Moral parliament, where I have a swathe of causes I care about and decision theories I consider plausible
Relatedly, self-coordination-like arguments; I got into EA from a global health angle and even though I’m more xrisk-pilled now, it seems good to respect my past wishes
Donating as a way of tracking & registering predictions of which charities will do well, similar to placing public bets on prediction markets. The information value gained from doing the exercise helps with making future donation allocations; or updating my model of what charities are good for future recs (for funding and for jobs)
Nah, the real prize is the karma & engagement we receive along the way!
Winners of the Manifund Essay Prize
Welcome to the EA Forum and thank you for posting this! I enjoy both Change.org and now Givedirectly. I agree with most of your points (and every one of your hot takes, I think!)
I’d push back a bit against “4. Use the index funds of giving.” One nitpick is that I’m not sure the analogy holds that well—charitable funds like Givewell and CG’s are more like mutual funds or hedge funds; you can’t actually passively index because there’s no simple metric of market cap to benchmark against. So implicitly, the choice of which fund to give to bakes in a bunch of worldview and effectiveness assumptions, and a lot of trust in the people running the charitable fund (unlike, say, VTI).
More broadly, I think that on the margin, there’s too much deference to charitable funds and too little “do your own research” in the EA space. (Though I understand that your original post on LinkedIn is angled for a wider audience, and there, I think Givewell—or Givedirectly! - is a great default rec).
The problem with charities/DAFs accepting pre-IPO stocks though is that they still need some way of liquidiating those stocks at the end of the day.
There’s also more exotic things that are possible, for a large enough donation size. A year ago before the Anthropic tenders were worked out, I had a proposal for lining up Anthropic employees & EA earn-to-give donors, and having them do a donation swap.
Some people (earn-to-give folks? banks?) may be willing to lend you money against your private/pre-IPO stock as well?
I think preparing for AI money is generally smart given Anthropic & OpenAI Foundation, though I don’t expect Ineffable specifically to have liquidity for at least a couple years.
It’s possible that there are some clever schemes that could allow David or others to start donating sooner (eg some liquidity at a raise, or borrowing against value of stock), but historically it’s not until IPO (and sometimes much later) before founders donate significant amounts.
soon! writing up some feedback on the winning essays and some reflections
I will consider this, thanks for the nudge!
Yes, Marcus Abramovitch and I put out this piece analyzing cost-effectiveness for AI safety youtubers specifically.
Manifund doesn’t have other pieces in the pipeline, but I would love for more work of this kind to exist, and I know other initiatives like https://grantmaking.ai/ are interested in finding qualified folks to do this kind of analysis at scale.
Thanks for asking!
My strong default prior is that most forprofits are good for the world, along the standard arguments: gains from trade, Paul Graham on wealth, the finding that corporations only keep 2.2% of value created
Moreover, I like when people who share my values start valuable companies, because they often spend that money on projects that are good. Mechanize doesn’t seem that different than Asana, or maybe Microsoft in this regard. Tamay has taken the GWWC pledge; I find Matthew’s writing (eg on AI rights) very informative.
Object-level, it seems like Mechanize mostly sells good code RL environments to Anthropic. Across the community, opinions on accelerating Anthropic capabilities are also mixed, but on balance I lean pro.
I personally benefit a lot from the good coding capabilities that Claude Code provides. This stage of AI/LLM development seems broadly good to me.
(Nb, my view wouldn’t change if they were mostly selling to OpenAI or GDM or something.)
On a minor level, some people view that leaving Epoch was somehow a betrayal of Epoch or the funding they received; this seems quite fake. I strongly support individuals’ rights to branch out and start new orgs. In any case, it seems like Epoch has continued to do well.
I included this story as a short anecdote about Marcus’s ability to spot talent, make active investments, and convince founders to take the leap, all of which I expect to transfer into helping start great AI x Animal orgs. I understand that different people in EA/AI safety have different takes about whether Mechanize specifically is good or bad—I happen to think good or at least neutral.
(And I take responsibility for any factual errors with this specific anecdote. Talking to Marcus just now, it seems like his main nudge was to convince Ege/Matthew/Tamay that the nonprofit structure was wrong for what they wanted to accomplish.)
Thank you for the retro! As we’re planning out our own incubator for for-profits that do good (Surplus), I really appreciated the section on “advice for other incubators”. We are indeed providing seed funding ($100k), and are focusing on software as a vertical (and specifically AI for epistemics, community infra, and informative websites), so I’m hoping that we’ll be able to improve on Founding to Give. It’s also a 3-mo IRL cohort in SF!