Appreciate this framework, @Gergő Gáspár . A few thoughts from where I sit at Successif:

(1) This framework could work just as well flipped around. I.e. talent uses it to decide how much of their own time to invest, and where.

(2) Having landed on similar factors within the team, a major challenge has been testing for them. E.g. LLM-polished applications make self-reported answers meant to indicate value alignment harder to assess. 

(3) I wonder about adding a fourth factor: situational constraints. I.e. Practicalities that cap someone's options regardless of how they score elsewhere. Should be easiest to measure, though unsure how hard to filter for these, as in some cases in seems to perpetuate structural inequalities. Examples:

- Visa/citizenship status: you can clear the calibration line and still be unable to transition due to geographic mobility barriers. 
- Willingness to relocate:  We've seen a recent wave of senior ops roles open up in AIS hubs like SF and London, but out of our top advisee talent I could recommend <10% because of a strict unwillingness to relocate. 
- Transition timeline: needing income in 1-3 months is a different case than a 6-12 month runway. 

A situational parameter I'm most uncertain (and curious!) about is whether people unemployed/on sabbatical putting in 30-40 hours a week cover meaningfully more ground than those working full-time with a few spare hours a month. 

As a career advisor at Successif, I've been thinking about this question, too, for those who do transition - but more from the angle of AIS team culture/organisational values alignment. Your socio-relational angle is an interesting take  I hadn't considered. I'd be interested to read your findings, and potentially integrate into my advising practice. 

Will you publish it here? Perhaps there's a mailing list you can add me to? [email protected] 

Hi Kariema, I appreciated the visual examples and hyperlinked cases of "human failure". Stories to share around a dinner table! 

I agree in principal that individuals have a responsibility to maintain oversight. However, I think the characterisation of "human failure" as an individual's problem (i.e. someone "not bothering to check") and the proposed solution of "us standing guard" is only a piece of a wider puzzle that necessitates identifying and addressing the complexity of situational/ structural factors (see fundamental attribution error). 

In each of the example mentioned, I'd be curious to know why the human failed? Moral imperative aside, what incentivised them to "fail" at the time? What factors can be introduced (on individual and structural levels) to incentivise the "human success" you so passionately advocate for? 

My big take-away is seeing fear itself as a policy-related variable, and that effective AI governance must consider emotional infrastructure alongside institutional infrastructure.

I'm left wondering how psychology concepts like self-determination theory scale across the individual/micro to the macro (e.g. collective action, institutional behavior, movement building, etc...).

On a meta-note: As a career advisor in this space, a common bottleneck I observe from mid-career professionals is deep uncertainty as to how non-technical experts can contribute to reducing AI catastrophic risk. I hope this work signals what multi-disciplinary thinking can bring to the space.

Similar curiosity, and similarly surprised! 🤓 

As an advisor at Successif, the question of how/to what extent AI safety fellowships serve as effective ramps for an AI safety career transition come up frequently with advisees. 

 I'm interested to see how your project develops, to support better informed decision-making.

As co-founder of SyDFAIS, an interactive visual map of the AI safety and AI development ecosystems, I'm tickled thinking how your data can feed into the visual map to facilitate sense-making and even imagining future scenarios across multiple filters (e.g. gender, location, fellowship focus, skill profile, etc) and for multiple stakeholder groups (funders, fellowship program designers, career transitioners, etc) . Will bookmark this @Christopher Clay, to revisit once our prototype goes public soon.

I coached Gergo on people leadership and project management for 6 months and can personally vouch for his dedication and transparency. 

From my current work at Successif (supporting mid-career professionals transitioning into AI risk reduction roles), I see Amplify's work as complementary to the broader ecosystem... addressing the top of the funnel (getting talented people aware of and engaged with EA/AIS) while organizations like ours work further down the pipeline on career transitions. 

+1 on the catch-22 @David_Moss describes. But might framing this as a "catch-22" unintentionally make the problem seem static and unsolvable? Bridgespan's research on field-building across 30+ fields identifies 'infrastructure' (i.e. connective tissue) as one of five critical characteristics for achieving population-level change and that these mature over 3 distinct phases. Perhaps what we're witnessing is a growing pain as the EA/AIS field transitions from the 'Forming' phase towards 'Evolving & Sustaining'...where the field has developed strong actors, a strong knowledge-base, and a (somewhat) shared agenda, but the funding mechanisms haven't yet matured to reliably support the intermediary infrastructure that enables coordination at scale.

A systems thinking approach might help with breakthroughs: What are the feedback loops keeping this pattern stuck? (e.g. funders see marketing as "not meeting the bar" → groups don't build marketing capacity → marketing continues to underperform → reinforcing funders' initial assessment). What leverage points could shift this? In the tobacco cessation field, the Campaign for Tobacco-Free Kids received sustained funding precisely because funders recognized infrastructure's multiplier effect across the entire field, not just individual programs.

Wishing you success with this funding round, Gergo, and/or with the rich learnings harvested along the way.

I appreciate your examples of how 'attitude' can be assessed at each stage of the application process. 

I'd be interested to hear perspectives from hiring managers or people ops in the space, especially in orgs that are scaling and where sub-optimal attitudes have wider implications on organisational culture and- ultimately- impact.

I wonder, William, what factors you think would incentivise or disincentive an org to integrate/test some of these ideas during a next hiring round?

Thank you for sharing, Christen. Trying to break into a space that seems to require the very experience you're seeking is tricky. Actually, your comment prompted me to realize some biases I'm likely operating under, which I've now included in the section 'What this List is Not'.

I'd think about "high-impact experience" in a CV differently. It doesn't have to come in the form of formal job titles. What hiring managers could find equally valuable is evidence that you understand the space and can contribute meaningfully which can be demonstrated in several ways.

In my own CV, I didn't have traditional AI governance experience either (I had "school principal" and "consultant"). After several months of 'journeying', though, I included:

• X# of conversations in the space
• Y# of volunteering/pro bono consulting projects (here is the doc I'd hyperlink, which of course started with only one project)
• Z# of high-impact events attended
• All relevant courses I'd completed 

(To see an example, visit my LinkedIn --> Experience --> Career transition)

This approach essentially "substituted" formal experience with what I call "acquiring context".

Regarding those 170 actions - let me break that down because it's less daunting than it sounds! Over one year, that's roughly 3 actions per week, or one every couple of days. I acknowledge that available time greatly affects this, especially while holding down a full-time job. 

I was never good at keeping a diary, but I found tracking actions helpful as a project management tool (I used RAG colors - red for rejected/door closed, amber for pending, green for accepted/success), with hyperlinks to easily retrieve previous applications or contacts, and finally as a way to motivate myself with small sense of progress (it's a game mechanic that works for me).

As for a more direct path - while there isn't really a "streamlined" route, there are definitely "conveyor belts" or "nodes" with high traffic under those three prongs: networking, small projects/pro bono consulting, and upskilling:

EAGs - I attended 2 during the year and found them (extremely) helpful for developing context and walking away with warm networks (I prefer the term 'relationships'), volunteering opportunities, etc. Not sure where you are in the world or what your capacity is, but the next couple months is EAG season.

Upskilling courses like BlueDot Impact (apart from the content) are positive market signals and connections to networks. The capstone projects were an opportunity for me to work on something directly in the space (e.g. milestone #10, #12, #14).

Other conveyer belts (alternatively: on-ramps?) include direct career transition support, such as 80k Hours (one-off career advising), Successif (long-term, relationship-focused career advising specifically in AI risk mitigation for professionals with any 5+ years experience), and the Impact Accelerator Program (6 week structured program within a cohort). AI Safety Collab (8 week course) and fellowships (e.g. GovAI; Arcadia Impact) are likely conveyer belts, although I did not do these myself. 

The whole point of my post is to show that you won't find a streamlined path, but to invite you to create your own path and take full advantage of your circle of control. And perhaps to sound less cliche - whatever time you're spending on career transition, consider redistributing to ~40% applications, ~20% deep networking, ~20% upskilling, ~20% small projects/volunteering/pro bono consulting. (i.e. a portfolio approach!). Then again, increasingly conscious of my stated biases.

Hi Tania! Your ops/strategy background is relevant for impact organizations - many of the challenges you're describing (sectoral gaps, positioning transferable skills) are covered in our recent post "Challenges from Career Transitions".

Your leadership and user-centered design experience could be valuable in impact orgs, though the transition often requires deep networking and strategic upskilling alongside applications - all with a 'winning or learning' mindset.

I went through a similar journey myself, which I wrote about in "To the Bat Mobile!! My Mid-Career Transition into AI Safety" - found that connecting authentically with people already doing the work was more valuable than cause-area expertise initially - although developing 'context' later in my journey proved essential.

Operations/strategy roles do seem open to career switchers in my experience. As a rough heuristic, worth noting that smaller orgs may expect more cause-specific familiarity since ops roles wear multiple hats, while larger orgs tend to have more specialized roles requiring less cause-specific knowledge (although not a hard rule!). Also, "ops" varies widely between organizations - always check the actual role description to see if it's focused on finance, HR, compliance, etc or combines everything.

All the best,
Moneer (Career Advisor at Succesif)