Comments
From 2022 to 2026, ML4Good has run 31 in-person bootcamps, and we have 600 alumni across the world. We select promising participants who have already been introduced to AI safety, and are in a position to move into competitive fellowships and AI safety roles. We’ve done this across 5 continents, mostly in places where there are few other AI safety programmes, and have had a major counterfactual impact in seeding new global communities.
We have just posted our impact assessment, which found that over 180 of our alumni now work in AI safety, or about 135 full-time equivalents. We estimate that about half of this is counterfactual. Across our total spending, we estimate a cost of around £9.3k for each counterfactual FTE.
When ML4Good started in 2022, very few people outside the Bay Area and the online communities of Effective Altruism and Less Wrong/Rationality had heard of AI safety. We started out as a volunteer-based team focused on getting technical people into technical AI safety roles and programmes.
The world has changed.
The overwhelmingly bad news is that we seem to be in the short timelines world. AI capabilities are advancing rapidly, and frontier companies may be close to reaching recursive self-improvement, without anywhere near the assurances we’d need to make sure that this technology will be developed safely.
The good news is that people are starting to wake up. A new pool of talented, motivated, and influential people around the world is getting interested in pivoting their careers or orienting their professional work towards AI safety and governance.
The field urgently needs many kinds of profiles: technical researchers, policy and governance experts, ops professionals, lawyers, biosecurity specialists, communicators, motivated generalists, and groups that few organisations are reaching yet.
At ML4Good, alongside our technical programmes, we now run in-person programmes for a much wider range of profiles. In our recent programme for lawyers, we targeted legal professionals in a position to shape policy and corporate practice in ways that the movement hasn’t yet considered. We've seen directly how they've oriented their work to safety-focused areas after the programme. We plan to keep making these bets, expanding into areas where the field has gaps, and catalysing the growth of new programmes.
For the next couple of months, we’ll be focusing on scaling the org, refining our strategy, and building connections with other orgs in the AI safety space and beyond. These are some things to expect from us in 2027:
We’ve had a few changes recently. Nia Gardner, who has been heroically leading ML4Good since 2024, is stepping back from day-to-day work to join the Board of Directors. Jack Stennett is now leading as the Director.
We’re currently working with a team of 3.6 FTE, and we’re planning to scale to 6 FTE by the end of 2026.
We're hiring for the following roles:
Permanent roles (applications close 15 November):
Contract role (applications close 1 November, reviewed on a rolling basis):
If you know someone who's a great fit, please recommend them here using this form. We will offer €200 reward for teachers, and €500 to any referrals we end up hiring. Find the referral award guide here.
We’re starting a global hiring round for teachers in Europe, Asia, Latin America, the US and elsewhere. If you’re working on AI safety, either in a full-time role or in a competitive fellowship, and think you’d make a great teacher, please apply: https://ml4good.org/jobs/ml4good-teacher/
Our teachers regularly tell us that teaching at ML4Good is the highlight of their year, and that it’s a welcome break from their day-to-day work in research or policy. Teachers have also written about how teaching at ML4Good is one of the best things you can do to build skills for other roles (e.g. research manager) in the AI safety space.
