In gentle of AI progress this 12 months, we’ve determined to cease posting senior-level and mid-career international well being, animal welfare, and local weather change roles on the 80,000 Hours job board. We are going to proceed to submit entry-level and junior roles, and we’ll hyperlink to different job boards that target these areas.
Why we’re not posting mid- and senior-level roles in these areas
AI progress this 12 months (Claude Opus 4.5, 4.6, and 4.7, GPT-5.3-Codex, and Claude Mythos, all of which represented sooner functionality progress than I anticipated) has satisfied me that the anticipated impression of labor on transformative AI (TAI) has grown dramatically, relative to that of labor on international well being, animal welfare, or local weather change (henceforth: non-TAI), and that we’re getting into an all-hands-on-deck state of affairs for TAI. In the meanwhile, our mission is greatest served by specializing in AI and roles associated to catastrophic AI-mediated dangers (e.g. biosecurity or AIxAnimals). It is because:
Now we have quick timelines to transformative AI: We predict it’s fairly believable that transformative AI will arrive within the subsequent 5 years. Whereas I was much less assured on this, the progress since Opus 4.5 has made me suppose it’s fairly doubtless that by 2040 we may have reached transformative AI.As a result of we’ve so little time, transformative AI going effectively “by default” now appears comparatively unlikely to me, which implies there may be a whole lot of work to be performed. AI security, biosecurity, and macrostrategy stay very uncared for.Situations the place AI capabilities ‘plateau’ now appear impossible. In such situations, the world wouldn’t change dramatically as a result of AI, and jobs in different areas would turn into comparatively extra vital. However at this level, it looks as if a mistake to not deal with worlds the place the traits of the previous few years proceed.
We intention to offer our customers with essentially the most promising roles for engaged on the world’s greatest issues. However in a world with imminent transformative AI, a very powerful intervention for human well being and animal welfare is ensuring that that transformation goes effectively. Thus, we’ve stopped recommending roles for knowledgeable customers on interventions that don’t interact with the approaching adjustments. We consider their expertise would, in expectation, go considerably additional in roles targeted on TAI.
This isn’t a cheerful replace for me. As I wrote throughout our strategic change in focus final 12 months, I hate that that is the timeline we’re in. And whereas we may proceed to submit these roles as a result of it’s what we’ve all the time performed, as a result of we don’t wish to lose a part of our viewers, or due to an obligation to the EA neighborhood, we might relatively act on our perception that we’re getting into an all-hands-on-deck state of affairs.
Why we’re nonetheless posting entry- and junior-level roles in non-TAI areas
Along with roles with excessive direct impression, we submit roles the place we consider folks can construct profession capital to deal with the world’s most urgent issues. We consider that in sure early-career non-TAI alternatives, folks can have impression on the world and construct expertise and habits of thoughts that might be onerous to seek out elsewhere (For instance, many international well being and animal welfare charities within the EA house put a premium on calibration and expected-value pondering to an unusual diploma; each are very worthwhile in TAI work.)
We’ll be making this extra clear on the job board so folks perceive our views. In a method, we take into consideration the first worth of those roles otherwise from the organisations that submit them as a result of we hope that individuals who take these roles will develop expertise to later work on TAI. Nevertheless, we consider that is higher than us not posting these roles in any respect for the organisations.
Why we’re nonetheless posting roles in different trigger areas
No adjustments are deliberate for our curation of roles in biosecurity, nuclear safety, constructing efficient altruism, or macrostrategy. We consider that these all have an vital half to play as highly effective AI allows dangerous actors, disrupts strategic stability, and creates new philosophical inquiries to grapple with. Additionally:
Biosecurity continues to look robustly good even with out the issues of AIxBio.Macrostrategy and constructing efficient altruism are vital for persevering with to probe the central questions of EA, together with “The place can we do essentially the most good?” Regardless that we’re dedicated to our present reply to that query, we predict it should proceed to be requested, in order that we are able to replace on future proof and concepts.
Some vital arguments and questions that we’ve thought-about
Non-TAI work stays rather more tractable.That is honest, however we predict this issue shouldn’t be sturdy sufficient to beat the elevated urgency and significance of TAI roles.
Match issues; some folks’s most impactful alternatives received’t contain TAI.An individual’s match for a job kind normally doesn’t depend upon its trigger space. All areas want operations, technique, communications, and so forth.
Additionally, we’ve all the time had an impression ‘bar’ for roles, which signifies that some folks’s private greatest alternatives can be excluded. It is a concrete draw back of getting a bar, however the upside of getting a bar is readability for our viewers about what we predict the very best choices are.
Received’t this shrink your viewers, leading to much less consideration to the trigger areas you’re now prioritising?That is attainable, however we predict it’s an appropriate danger if it lets us act on our true beliefs. Moreover, our viewers is already strongly AI-focused. Roughly 88% of our job board customers since our first pivot final 12 months have clicked both solely or largely TAI-relevant roles, 4% have clicked a stability, and eight% have clicked solely or largely non-TAI roles. We anticipate the AI-focused viewers to develop alongside public curiosity in AI, and we don’t anticipate engagement to considerably decline with this variation. Even when it did, that might not be a ample purpose to proceed posting roles that don’t appear as promising to us.
Why hassle posting non-TAI roles while you’ll have even fewer now?We proceed to suppose that non-TAI entry- and junior-level roles are above our bar due to the profession capital they supply.Moreover, this variation hasn’t led to a dramatic drop within the variety of non-TAI roles. Whereas we’ve eliminated senior- and mid-career roles in these areas, we’ll be barely increasing our standards for early-career roles, which is able to considerably make up for it.
Will this put you out of sync with the remainder of 80,000 Hours?The job board has been extra pluralistic than the remainder of 80,000 Hours since final 12 months’s pivot. This transformation will put us extra in step with our different programmes (akin to advising), that are strongly TAI-focused.
Different job boards
We predict there are cheap views inside which non-TAI roles may very well be simply as impactful as TAI-relevant roles (e.g. Marcus Davis’s arguments for epistemic humility). In different phrases, we may very well be unsuitable. Thus, we’re comfortable to advocate different job boards which focus extra on these roles, and now characteristic hyperlinks to those job boards extra prominently on our personal, in order that individuals who come to us on the lookout for non-TAI roles can nonetheless discover what they’re on the lookout for. These job boards are:
Now we have been coordinating with these different job boards to make sure they’ll proceed supporting organisations we’ve supported up to now.
Appreciation
I do know that that is unwelcome information for some in our viewers and for a lot of organisations that we’ve supported for years, and I’m sorry. We’re making our greatest guess concerning the biggest risks of the subsequent decade. It’s my expensive hope that we’re unsuitable.
I made this determination as 80,000 Hours’ job board supervisor, however my sturdy appreciation goes out to a lot of my colleagues for his or her contributions to and debate on this determination, notably Huon, Arden, Jess, Bella, Jenna, and Alice. Thanks additionally exit to Dylan and Thane at Most likely Good, Agnes at CEA, and Ana at AAC for his or her spirit of collaboration. Thanks are additionally as a result of Nina Friedrich, Zach Robinson, and Gergő Gáspár for his or her ideas.
