The current Long Term Future Fund (LTFF) fund managers and I have decided to step back from our work on the LTFF. Because we believe LTFF donors trusted the fund managers to ensure that the funds would be used in line with the purposes of their donation, we've decided the right move is to close the fund.
While LTFF is closing, note that EA Funds has launched a new fund, called the Transformative AI Fund (TAIF), with a large overlap in scope, funding early-stage projects that reduce risks from powerful AI systems, led by Lowe Lundin. It is open for applications and donations now. Some LTFF Fund managers plan to continue part-time grantmaking via the AI Risk Mitigation Fund and Lightcone Commons.
LTFF fund managers have discussed what to do with the remaining balance of the LTFF. We see our duty to LTFF donors as maximising the expected impact of the money they entrusted to us, within the scope of the fund. Closing a fund involves trade-offs in how to honour that.
With that in mind, we have decided to transfer the remaining balance to [new fund], structured as follows:
Up to three-quarters will be distributed under the direct oversight of LTFF fund managers, primarily through the first Lightcone Commons funding round. Lowe will also make recommendations as part of this process, but LTFF fund managers will ultimately decide on the final funding allocation. The remainder will go to seed TAIF’s first independent grants.
With that in mind, we have decided to spend down the remaining funds as follows:
- ¾ to distribute to existing LTFF applications and longtermist projects (~$2.8M)
- ¼ to the new CEA-housed AI safety-focused fund as a seed grant. (~$0.9M)
Why are we moving on from the LTFF?
Over the last few years, the pace of AI progress has been very fast, and a lot has changed in the AI safety community. When the LTFF started in 2017, there were few other funders making grants in the AI safety space, but over the last few years other grantmakers have set up high-volume grantmaking programs in technical AI safety, career transition grants, policy, and fieldbuilding. There is significant interest from other funders in how to deploy billions of dollars in philanthropic capital from AI company employees, and we regularly advise new funds entering the space. In many ways, there hasn’t been a more exciting time to get involved with grantmaking!
That said, I think there are several reasons why transitioning from the Long-Term Future Fund is the right choice for other fund managers and me. Note that I am writing in my own voice here, and other fund managers may agree or disagree with various parts of this.
Situational Awareness and Quickly Reorienting
There has been much discussion about the need for more “grantmaking capacity” in AI safety in light of a windfall of AI money for philanthropy. When I think about addressing the need for deploying more funding cost-effectively, I don’t think that “evaluator hours” describes the largest bottleneck for distributing billions of dollars effectively. I’m more compelled by increasing the AI safety community’s ability to notice underrated and important considerations (and then act decisively and effectively).
Grantmaking within CEA greatly benefits from focused, dedicated capacity - particularly for crunching through the kinds of small grants the LTFF is most focused on and coordinating with operations teams. Unfortunately, I don’t think that full-time grantmaking is a great fit for developing the kinds of “deep” takes that could significantly level up our grantmaking.
- You spend relatively little time getting in contact with the reality of building organisations and object-level technical work.
- It’s rare for grantmakers to have “deep” models in any area without many years of prior experience, which makes it hard to build conviction in highly promising but overlooked or “low status” ideas.
- It’s hard to quickly reorient unless you make a conscious effort to do so because of past commitments to the community and having to regularly actively solicit information rather than being forced to regularly confront “the abyss”.
To address this, part of the thesis of the LTFF has been to largely rely on the judgement of people who spend most of their time doing the kind of work we want our grantees to do (e.g., running organisations, fieldbuilding, technical AI safety research, and AI policy), and a minority of their time grantmaking.
Competing Opportunities
Other fund managers and I have other valuable things we can do with our time. Often these competing opportunities have a lot of synergy with the kind of grantmaking the LTFF has been doing but feel more impactful than direct grantmaking. For example, Oliver Habryka recently started Lightcone Commons, which will hopefully result in many early-stage ambitious AI safety projects being started or accelerated.
Institutional Friction
Grantmaking is a legally complex endeavour, and many of the grants that we have historically been most excited about making have also created the most legal risk for the Centre for Effective Altruism (and associated projects). Most of what LTFF funded sat comfortably within what CEA can do. But a meaningful share of the grants I was most excited about were unusual in ways that made them legally complex (e.g. grants to individuals rather than organisations, to for-profits, or to work adjacent to policy and non-partisan advocacy) or created reputational risks for CEA and EA.
Many of those grants took disproportionate time and created disproportionate risk for CEA, and over time I found myself shaping what I was willing to consider around CEA’s operational realities. In my view, this is a reasonable trade for CEA to make, it just isn't the trade I want to make, which is a matter of preference rather than a problem the new fund inherits. My understanding is that other LTFF Fund managers have similar preferences to me.
Why Is EA Funds Starting a New Fund Making AI Safety Grants?
So why are we closing LTFF rather than simply handing the name to the new team? The main reason is that I think LTFF's track record is mostly a reflection of particular people's judgement, and a fund carrying the same name with an entirely different team would start out with a reputation that for better or worse the new team hasn't yet earned. A different name pushes donors and applicants to orient to the fund as a new effort, and lets Lowe build a record and be judged on it.
What will happen with remaining funds?
As mentioned above, we expect to distribute 75% of our remaining funds with the same team to “traditional” LTFF projects with 25% going to TAIF, the new EA Funds fund.
I think this has a bunch of great properties:
- We accelerate a new grantmaking fund and meta funding process that will both hopefully be productive grantmaking entities for years to come.
- Funding is spent in ways that I believe are faithful to the beliefs and preferences of our donors.
- We don’t need to significantly lower our bar to spend down remaining funds so that we can wrap up.
- We create some brand separation between the new TAIF and the LTFF, which I see as practically different institutions.
Logistically, we expect to spend about ¼ of the budget as a seed grant to the new CEA-led fund using our existing grantmaking processes at CEA, and the rest through
- Our normal process for existing applicants.
- participation in Lightcone Commons, which is a new funding platform that fund member Oliver Habryka just started, which will handle a bunch of the logistics and funder cross-coordination associated with disbursing the rest of our fund balance.
Why Lightcone Commons?
Lightcone Commons runs on the S-Process, the algorithm developed for the Survival and Flourishing Fund and Lightspeed Grants, which has been used to distribute over $100M in grants. Evaluators submit a priority order over projects, funders choose which evaluators to defer to and in what proportion, and when several funders want to fund the same thing they split the cost in proportion to how much they are giving away. The first round expects to disburse $15M to $25M, with participating funders including Jaan Tallinn, Dustin Moskovitz, the ARM Fund and several anonymous donors, and rounds thereafter roughly every three months with rolling applications.
Almost all Long-Term Future Fund applications have been imported into Lightcone Commons (where we were given permission to do so).
This is comfortably within LTFF's scope. The bulk of our funding will be aimed at work on shaping the development of advanced AI; applicants do not need 501(c)(3) status; grants can go to individuals, for-profits, 501(c)(4)s, and non-US organisations; and live LTFF applications (where permission for sharing were given) have been imported into the round, so much of the pool is made up of exactly the sort of small, early-stage, hard-to-fund projects LTFF has specialised in.
Oliver, I, and potentially other LTFF staff will participate in the first round, so routing this money through the Commons means the remaining LTFF balance is still allocated in part by the judgement of the LTFF fund managers at the time the donors made their donation. To avoid any perception of conflict of interest, Lightcone has agreed to waive its administrative fees on this transaction, so the 3% platform fee and 2% evaluator fee will not be charged on the LTFF contribution.
Why Support the New CEA AI Focussed Fund?
One thing LTFF provided that I would like to preserve is a well known place where individuals with promising early-stage projects can apply at any time, without needing an introduction. The new fund (TAIF) will offer roughly that: an open request for proposals, rolling decisions, and the ability to make small grants to individuals and early stage organisations.
A new fund starting from a zero balance would have to fundraise before it could commit to fund any projects and start generating its own track record. Providing a seed grant means the fund can hit the ground running. I plan to act as an advisor to the new fund, which I hope will help carry over some of what we have learned working on the LTFF over the years.
What do existing LTFF applicants need to do?
- If you have an outstanding application to LTFF and you have allowed for your application to be shared with other funders, you should have been notified that your application was shared with Lightcone Commons, and you do not need to reapply to Lightcone Commons.
- Live LTFF applications are also still being considered for the portion of our remaining funds going to existing applicants, and we will get back to you either way.
- Where an application is within the new fund's scope, and we think it is promising, we may also pass it to Lowe's team for consideration (if you gave us permission to pass on applications to other funds), including in cases where we have decided not to fund it ourselves.
- If we transfer your application, we will tell you directly at the time, so you will always know where it has gone.
- We’ve also sent applications where permission was given to share with other funders to Lightcone Commons - you should have already received an email from Lightcone if that applies to you.
