Mark Aiken

Karma: 31

Mark Aiken 27 Aug 2026 13:21 UTC
2 points
0 ∶ 0
on: How to Properly Reject AI Safety Fellowship Applicants: Lessons from Rejection Letters from Top Organizations
Great writeup. This is also a really positive example of the kind of work that Gen Stream can produce by getting someone to focus on a specific issue for a short sprint, and publish the analysis. So kudos to you and to Gen Stream for this initiative.

I want to acknowledge that managing a recruitment process is a huge amount of work, and giving detailed feedback is an additional workload on top of this. Not all organisations will have capacity to do give targeted feedback, even if they wanted to. There are also some risks of giving targeted feedback in some jurisdictions—eg if your feedback indicates that the employer has unlawfully discriminated against the candidate.

However, the further candidates advance and invest in a selection process, the more they would benefit from getting targeted feedback. Feedback can be provided by stage as you’ve explained (candidates who reach the final round might get individual feedback, candidates who do not pass the first screening might only get general feedback). I also agree that feedback can be misplaced—if a candidate reaches a late round, they are probably well beyond the stage of BlueDot and 80,000 hours and referring these candidates to resources that they have already advanced beyond doesnt seem very helpful.

I’d be interested to see examples of good feedback if the organisations you’ve chosen decide to post them. From previous high volume experience, I’ve seen some organisations give effective feedback by describing what strong candidates did vs weaker candidates—eg strong candidates could clearly describe A,B and C and apply this knowledge to the scenario to identify risks, impacts, sequencing. Weaker candidates identified 2 or less of the key issues; did not link the issues to an impact and / or did not clearly describe a viable sequence. By showing candidates what good and bad look like, you invite them to find their own opportunities for improvement.

I also think there are more opportunities for common recruitment processes and recruitment for shared rosters across organisations. I know this is not easy to organise, but it’s common practice in some industries. Often organisations are recruiting for very similar competencies (rightly or wrongly), and it’s a significant workload on both hiring organisations and applicants to repeatedly have full hiring processes for common roles.

Mark Aiken 2 Aug 2026 18:32 UTC
12 points
3 ∶ 0
on: What if the third wave is a puddle?
Thanks Morgan, this is a valuable perspective.

One observation from an adjacent world: absorption capacity is treated here as a new and poorly understood problem, but it’s a mature discipline in the multilateral, development and humanitarian ecosystems. Surge absorption has been studied and managed there for decades, with identified tools and processes—eg absorptive-capacity assessments before scale-up, phased disbursement tied to implementation rates, pooled funds that spread money across implementers when no single organisation can absorb it, and monitoring systems that flag when spending outruns capacity.
The failure modes you list (unnatural growth, rushed strategy, hiring ahead of funding that doesn’t materialise, and plus, corruption) are also well documented there, from post-tsunami aid to large global health scale-ups. We know from experience that absorption fails at the organisational level before it fails at the field level; phased growth outperforms step-change growth; and reversibility (“two-way doors”) is exactly the right design principle.
So I’d add one suggestion to your four: the sector doesn’t need to derive this knowledge from first principles. There are many practitioners with extensive (decades-long) of experience planning, budgeting and disbursing at the scale you are concerned about. As one example, the budget of the Syria Humanitarian Needs and Response plan for 2026 is $2.9b, covering health, education, food security, water and sanitation. The experience of these practitioners who have disbursed at billion-dollar scale through these exact dynamics is currently almost invisible to this ecosystem. Importing that expertise seems cheaper than re-learning the hard lessons.

https://reliefweb.int/attachments/28e3e5cb-a476-44ee-9cb7-f69625bdf31a/syrian_arab_republic_2026_humanitarian_needs_and_response_plan.pdf

Mark Aiken 22 Jul 2026 19:28 UTC
2 points
0 ∶ 0
on: A Complete Overview of Our AI Safety Talent Pipelines (+ Interactive Visual!)
Well done Roman! The post is great (and important!). And I love visual storytelling like this—there should be more of it in the AI safety field. This Hans Rosling video is one of my favourite examples of data visualisation / storytelling, in case you’re not familiar with it already:
I think the pipeline (or funnel) is a a solid analogy—we are trying to direct people towards impactful contributions and impactful careers. And that pipeline is very leaky—there are a lot of people who are interested in contributing to fields like AI safety, but arent able to navigate the process. Some of the reasons you’ve mentioned in your article. Additionally, if hiring organisations get 300 to 500 applicants for each position, they can choose the top candidate, which potentially means that many similarly capable candidates are not selected. Some of those candidates will eventually get discouraged and find work in other fields—not because they were not passionate about making an impact, but because at some point they have done as much as they can, and they need a job. This is not the fault of the candidate—this is an employment marketplace failure, where people who have skills and want to work are not able to be matched to a role in what ought to be a rapidly scaling ecosystem.

Mark Aiken 22 Jul 2026 18:27 UTC
3 points
1 ∶ 0
in reply to: Josh Davidoff’s comment on: A Complete Overview of Our AI Safety Talent Pipelines (+ Interactive Visual!)
To give credit where credit is due, I have seen some organisations starting to provide rejection letters (at least at first round) which point candidates towards existing resources and courses that could help them gain more context. I agree this is a good practice and should be encouraged.

Mark Aiken 17 Jul 2026 10:06 UTC
3 points
1 ∶ 0
on: New talent is coming to AI safety. What happens next?
Well done, this is a valuable and thoughtful piece.
I agree that a lot of talent appears to be left on the table currently. There’s a high workload for hiring organisations, who need to sort through increasing numbers of applications (including AI prepared submissions). And there’s a high workload for applicants, who might collectively spend thousands of hours applying for a single post.
It seems there are worthwhile ideas on how hiring practices could be made more efficient—standardising taxonomies (what do we call positions and functions), establishing competency frameworks for common roles and tasks, standardised terms of reference for common roles, addressing challenges of AI use in applications etc. Through to more ambitious approaches, like running joint recruitments, establishing rosters of pre-qualified candidates, and expanding lower risk / lower commitment ways for organisations and applicants to test their fit.
It’s great that organisations can be selective and appoint the top 1% of talented applicants. But if the field needs to grow to meet urgent timelines and risks, it’s important to think about other ways that we can do things, while still preserving the culture and values. I think there are models and lessons learned from other industries, and some of these could be valuable for AI safety organisations.
A question—I understand there’s an operations group for colleagues working on Ops in EA organisations. Is there any HR working group to look at some of these HR challenges across the field, rather than taking an organisation by organisation approach?

Mark Aiken 3 Jul 2026 12:09 UTC
1 point
0 ∶ 0
in reply to: Weronika Żurek 🔸’s comment on: Talent Constraints in AI Safety: What We Know and What We Don’t
I’m with Racoocoonie and Claude on this one. It’s astounding to argue there are no pools of candidates outside the AI safety community that have experience in fieldbuilding, apart from EA community builders.
The environmental movement has decades(++) of experience in fieldbuilding. The Sierra Club was founded 130 years ago for environmental advocacy—it has chapters in every US state and around 4m members. There are social justice movements, LGBTQI movements, religious movements, labour movements, voter registration movements, health movements and development networks, each with decades of fieldbuilding experience. There are tens of thousands of people working in fieldbuilding and organisation roles within these movements.
All of these people are part of different networks (or ‘pools’, if you prefer). Many of these networks are also present on platforms like linkedin.
However, this is a very niche background, and people who have it are very difficult to find
Honestly, I think it’s not so niche. If it’s so hard to find people then perhaps we’re not looking in the right places?
The other alternative is that we’re quite happy with the pool of people we already have and we dont need to try harder to bring in other fieldbuilders. If that’s the case, let’s not say it’s because those people dont exist, please.

Mark Aiken 11 Jun 2026 20:49 UTC
1 point
0 ∶ 0
in reply to: Ben_West🔸’s comment on: Experience of a ‘generalist’ transitioning into AI safety in 2026
1. Title the role “Senior [whatever].” I think this is ok, but in many fields “senior” is a synonym for “old”, so this title causes talented young people to not apply (and untalented old people to apply).
  I think ‘specialist’ works as a title in this situation, without invoking issues of age—eg Finance specialist, Comms specialist etc.
For me TOR are important, and I wonder sometimes if enough time is spent by hiring organisations (including commercial organisations) in defining what they want out of roles. I see Chief of Staff roles in some organisations that are quite strategic, and in other organisations that are essentially an admin assistant. I appreciate defining responsibilities and tasks is especially tricky in startups, where the roles are harder to define, and you might want a true generalist who can do a bit of everything.
If an organisation advertises for a HR manager, and talks about how they are scaling up and rapidly recruiting, I might apply—I’ve done end to end recruitment roles, performance management, HR policy etc. If the organisation advertises for a ‘HR specialist’, and talks about recruitment, setting comp, HR compliance across multiple jurisdictions, I’ve got a better idea what they’re looking for and I’m going to screen myself out by not applying because I dont have all that expertise.
I dont think I’m an ‘untalented old person’ in the first scenario—the organisational asked for something that I’ve done before, so I apply based on the information that was available. But if the organisation asks for A but really intends B, then everyones’ time has been wasted.
5. Ask for referrals from people who I know well enough that I can effectively say “highly skilled generalist” and they will apply that criterion in a way that I would endorse. This is good but means I don’t hire from outside my circle.
If everyone hires from within their existing circles then the community doesnt grow—it’s just musical chairs with the same people swapping between same organisations. If the thesis is that the community needs to grow to be able to (1) effectively respond to urgent AI safety challenges and (2) effectively absorb and use anticipated increased levels of funding, then to my mind looking at recruitment practices is an important part of the solution.

Mark Aiken 10 Jun 2026 18:43 UTC
1 point
0 ∶ 0
in reply to: Emily Marais’s comment on: Experience of a ‘generalist’ transitioning into AI safety in 2026
I think alignment -is- the right word.
And I think it’s not necessarily about people’s ability to do work, but people’s perceived commitment to ‘the cause’ and their perceived ‘moral fit’ with the organisation.
I think there’s a higher importance placed on this by some organisations, and for some roles. But it seems that some degree of ‘alignment’ is required to be recruited to almost all roles in the EA eco-system.
I’m not taking a stance on whether this is good or bad—people can rightly have opinions both ways on this I think. Just making an observation that it’s a factor.
I’m also going to refer to this post that is based on ‘context’ being a missing criteria for external candidates, but I think this concept includes alignment.

https://forum.effectivealtruism.org/posts/b82SLXwEHRCs3TFJA/why-experienced-professionals-fail-to-land-high-impact-roles

Mark Aiken 10 Jun 2026 9:36 UTC
8 points
5 ∶ 0
on: Experience of a ‘generalist’ transitioning into AI safety in 2026
Thanks for sharing your experience Emily! I think your experience is consistent with many of the colleagues I’ve been speaking with as I go through bluedot courses, CEA bootcamps etc. People are interested in learning more about AI safety, they want to gain context, but the pathways for generalists are not so clear or well developed. I’ve spoken to mid-career and senior level people who would love to contribute, but its hard for them to find entry points after the initial courses and fellowships. I expect that EA is missing out on some interested and committed people, who cant figure out how to navigate networks to find opportunities (or who give up after a year or two of trying).
I appreciate that EA and AI safety recruiting is done in a high-trust context, and there’s a premium placed on alignment and evidence like public writing. I also think that working or volunteering with EA / AI safety organisations is probably the fastest way for people to advance their understanding, once they’ve reached some baseline level.
Similar to the ‘freelancers for good’ idea, I wonder if there’s a way for recruiting organisations to ‘projectise’ some of their generalist work a bit more, as an opportunity for recruiting organisations to get some work done, but with a lower level of commitment required in case the recruit isnt a good fit. For example, recruit a short term comms person to develop a communications strategy and brand assets, recruit a short-term HR person to streamline recruitment processes and create banks of exam scenarios, recruit a short-term operations person to look at how systems can be streamlined, bring in a short-term legal person to review templates for legal agreements etc etc.
These are just examples, and different organisations will have different needs. But it seems to me that by opening up these kinds of short-term / lower-commitment roles for external candidates, it would help grow the EA / AI safety field, while helping interested people build their EA understanding. At the end of a short-term assignment, the recruiting organisation ideally has some useful product, the candidate has a new work product they can talk about, and both the organisation and the candidate would both have a clearer sense of whether the ‘fit’ was right for them.

Mark Aiken 20 May 2026 20:18 UTC
2 points
0 ∶ 0
in reply to: Sophie Kim’s comment on: AI safety is extremely bottlenecked on grantmakers
Hey Sophie,
I have quite a lot (decades) of grant making and grant administration experience.
Happy to brainstorm with you if you’re interested in pulling together an introductory level course etc. Send me a DM :)

Mark Aiken 9 May 2026 16:20 UTC
2 points
0 ∶ 0
in reply to: huw’s comment on: A brief update on EA Funds, and a recruitment announcement: Head of the EA Infrastructure Fund
Adding to this—people with mental illnesses in developing countries are often stigmatised and shunned by their families, and at worst imprisoned. They are imprisoned due to (1) public order offences (ie being disruptive in public) and (2) a lack of other facilities to accommodate them long term (ie hospital facilities or mental health programmes).
There is a lot that could be done relatively cheaply if this was taken up as a priority.

Mark Aiken 9 May 2026 11:42 UTC
1 point
0 ∶ 0
in reply to: David_Moss’s comment on: Some surprising hiring practices I follow (as a hiring manager and grantmaker in EA)
I fully agree.
My understanding of most structured interview formats is that follow up or clarifying questions are still expected and encouraged. There’s limited value in robotically sticking to a list of identical questions, and missing out on the opportunity to get additional information with a follow-up question.
I find the best interviews feel like a structured conversation. There’s interaction between the panel and the candidate, because that’s how we really interact when we meet together. There are efforts made to help the candidate feel relaxed and comfortable, and to value the experiences that they share with the panel. I cover the same questions in the same order with the candidate, but we might spend longer on some questions than others, depending on the background and strengths / weakness of the candidate.

Mark Aiken 8 May 2026 12:26 UTC
2 points
0 ∶ 0
in reply to: Agustín Covarrubias 🔸’s comment on: AI Safety’s Biggest Talent Gap Isn’t Researchers. It’s Generalists.
|> If you’re a program manager in a small org, you’re basically making highly strategic judgment calls on a weekly basis, and when people don’t have strong mental models around AI safety, they tend to make the wrong ones. I think at present the majority of demand for soft ops roles in the ecosystem looks like this: they’re roles that require one to make frequent judgment calls that benefit highly from field-specific context and a solid internalization of our priorities.
Firstly, I agree that in smaller organisations people do a bit of many things, and having some technical background is helpful.
The way you describe the programme manager role sounds to me like a research manager type position. These positions are most likely to need a technical background. However, there could equally be programme manager roles which dont require technical expertise.
For example, if I’m managing the development of a training course, I hire a subject matter expert to develop the content. I’m managing timelines, I’m managing staffing, I’m managing budgets, I’m managing quality—but I dont need subject matter expertise to do this management role.
Even if I have some subject matter expertise, this expertise might not be fully transferable—I can have deep knowledge about using Excel, but if I’m overseeing training course development on how to design a more efficient electric car battery, my Excel subject matter expertise might be mostly irrelevant.
Now you could hire a subject matter expert to be this programme manager to oversee the development of these training courses. But they might not have the background or skills for this role, because—they are a subject matter expert, not a programme management expert.
More broadly—I dont want my research experts to following up on visas, or drafting a press release, or doing graphic design for an event invitation, or trying to work out which politician would give them a sympathetic hearing, or how much cash flow we need for the next month. There are specialists who have this expertise already, and it seems more efficient to work with these specialists where it makes sense, instead of overweighting on prior AI safety technical expertise.
> There are important exceptions to this, especially at bigger orgs, where soft ops roles can be more specialized and therefore there’s less of a need for a high degree of context and deep mission-alignment (for example, I suspect many soft ops roles at Coefficient Giving are like this).
I agree, I think that larger organisations have a greater capacity to carry specialised non-technical expertise. But I think this specialist expertise could also be available to smaller organisations, perhaps on a fractional basis? If the organisations saw it as being useful. And perhaps if there was a better way of matching expertise to smaller organisations?
I also wonder if there is a potential chicken and egg argument there - ie, that organisations that are willing to accept specialist support (perhaps starting with finance, HR, operations?) are more likely to be stable and successful and grow into a larger organisation? Or conversely, if you have a sole founder and rely on technical researchers for everything, that this kind of structure can place limits on the growth and sustainability of the organisation.

Mark Aiken 8 May 2026 11:58 UTC
5 points
1 ∶ 0
in reply to: Karen Maria Alston’s comment on: AI Safety’s Biggest Talent Gap Isn’t Researchers. It’s Generalists.
Hi Karen,
I made an account here so I can post this comment.
I wanted to thank you for raising these issues. Similar to you, I’m a senior professional. I’m part of Bluedot and CEA cohorts which also include other senior professionals (20-30 years experience). These professionals (and myself) have already had impactful careers, and are looking to pivot into new impactful roles where our skills and experience can be applied.
With my personal background, it would be much easier (and probably better financially) for me to pivot into corporate AI Governance rather than AI safety. I’d prefer not to do this since it seems less impactful.
I also find the distinction in these discussions between technical / researcher roles and ‘generalist’ (ie non-technical) roles to be an interesting framing. It seems to me that there are other ways of assessing mission alignment, and there is a risk of overweighting on technical expertise, as HannahGB questions below.
To give a concrete (but invented) example:
Scenario 1: Lets imagine that we have an AI safety organisation doing wonderful research. The researchers take their results, and present at a technical conference. Attendees are very impressed, and this inspires other researchers to continue to develop the research pathway.
Scenario 2: Lets imagine the same research at a different AI safety organisation. The researcher briefs the executive leadership of the organisation on their results. The organisation works with their policy lead to tease out the implications of the research results. The policy and communications leads work together to define an advocacy position and their proposals, supported by a strong briefing deck with talking points that are understandable by a semi-technical government audience. A lawyer helps prepare zero draft text for inclusion in regulation. The policy lead links the organisation with regulators from their network, to present the research and advocate for improved governance. The Comms lead arranges publicity and advocacy for the discussion through their comms networks. And the researcher still gets to present the research at a technical conference.
It seems to me like there’s a lot of Scenario 1 work happening in the field. And then some separate organisations working on AI governance and oversight from a legislative and policy perspective. My question is whether there is also opportunity for more of the Scenario 2 type approach?