Open Phil is seeking bilingual people to help translate web content related to AI safety, global catastrophic risks, effective altruism and adjacent topics into non-English languages

Eli Rose23 Aug 2022 23:52 UTC
111 points
37 comments8 min readEA link

[Read this first] Update to Open Philanthropy’s Translation Program (Edit Nov 7, 2023)

TL;DR: We’re moving away from part-time translation grants, and are seeking applications from individuals or teams aiming to work on translations full-time. If you want to apply for these types of projects, please do so here.

We’ve decided to discontinue the individual translation grant program in its current form. Previously, we invited applications from individuals doing or managing translations part-time, most of them focused on one language. We made around 40 grants totaling around $950,000, and we celebrate all the progress that has been made. Thanks to everyone who made this happen!

Due to the limited capacity of our team, we unfortunately have to move away from this model and are now aiming to fund translation efforts of a bigger scale. We’d be interested in receiving applications from individuals or teams who are keen to work on translations full-time and to translate a significant amount of content. We’re open to different approaches to this, but roughly we expect it to involve:

  • Contracting with translators (vetting and paying them)

  • Ensuring the quality of the translated content (this could be done in various ways)

  • Ideally, but not necessarily, distributing of the translated material (getting it published in such a way that it reaches its target audience effectively)

What we would be keen to see:

  • Somebody who is excited to work on translations full-time for at least half a year, and potentially longer if it goes well. They intend to set up an entity, possibly together with another person, via which they will pay contractors to do translation work. They have thought about how to vet possible translators and how to check the quality of finished translations. They may have thought about where it makes sense to publish the translations.

What we’re moving away from:

  • Somebody who would mostly prefer to translate themselves (either full-time or part-time), and who doesn’t intend to oversee other translators.

We accept applications via our general application form here. Please note that the previous translation application form is now closed.

___

Summary

I work on Open Phil’s Global Catastrophic Risks Capacity Building team. Our broad goal is to grow and support the pool of people who are well-positioned to work on global catastrophic risks. One thing we’re interested in is making it the case that more people are exposed to high-quality content — books, blog posts, papers, podcasts, videos, articles, etc. — on relevant topics, like AI safety, global catastrophic risks, effective altruism, longtermism, rationality, etc. (This includes object-level content on cause areas like AI alignment and governance, biosecurity, global priorities, etc.)

But there’s a lot of high-quality content of this sort that’s currently only available in English or a few other world languages. This significantly limits how many people the content can reach. For this reason, Open Phil is interested in funding work that leads to this kind of content getting translated from English into other languages.

We’re looking for people who are able to work full- or at least half-time (at least 20 hours a week), either a) translating content themselves, or b) hiring non-EA professional translators and reviewing their work for quality.

Apply here if you’re interested in paid opportunities, part- or full-time, to help get content translated.

[Edit Mar 22, 2023: Changed the required commitment from 5 hours to at least 20 hours per week, to reflect the types of projects we’re most excited about at present.]

FAQ

Why is Open Phil excited about translation work?

Translation work looks to me like an unusually concrete task aimed directly at a serious bottleneck in community-building. It can also be done remotely and on one’s own schedule, unlike other types of community-building work (e.g. running meetup groups).

One simple way of thinking about the impact of translation work is by thinking about yourself as a miniature copy of the author. For example, the post On Caring took a certain number of hours for its author (Nate Soares) to write and has accrued a certain number of readers over time, which led to a certain amount of impact (which I think was positive). Currently, I’m not aware of any translation of this post into French. If you were to cause it to be translated into French, it would start accruing readers (probably fewer than it did in English), and after a while it’d have had some amount of impact, which would be some fraction of the impact Nate had in the original writing. In this scenario, you effectively spent some of your time acting as a mini-Nate.

The “miniature version of the author” framing might make translation work sound like an unbelievably good use of time. Some important things this framing doesn’t highlight:

  • The resulting translated work still needs to somehow get distributed to readers (which won’t happen automatically).

  • The impact of a translation may be higher or lower than its raw number of readers would suggest.

    • On the “lower” side: it is unfortunately, currently more difficult for people who don’t speak English to get involved in the EA community, for a variety of reasons (location of jobs, location of events, etc.). So readers of translated work may struggle to turn their interest into impact.

    • On the “higher” side: There’s a lot less EA-related content in non-English languages. This means that someone who reads your translated work might have been less likely to encounter similar ideas elsewhere, which gives you more counterfactual “credit” for whatever impact they go on to have.

But overall, I like how this framing highlights some core parts of why translation work is highly valuable per unit time, according to me:

  • Translation is easier than original creation, so you can piggyback off the best authors’ work without yourself needing to be among EA’s best writers.

  • You have the benefit of hindsight, so you can choose to “write” (i.e. translate) just the 1 out of 10 or the 1 out of 100 pieces that turned out the best.

(I think it’s tough to make general comparisons between translation work and other community-building activities in non-English-speaking countries, such as running groups or events. Which is more impactful depends on personal fit and other details.)

Who should apply?

You should apply if:

  • You enjoy EA/​EA-adjacent content and want to help share it with others.

  • You’re fluent in a non-English language.

  • You’re willing to do (paid) half-time or more (at least 20 hours a week) on translation over at least the next six months.

    • This could include doing the translation yourself, or supervising professional translators paid for by Open Phil (see below for an explanation of what this might look like). If you think you might be willing to do the latter option, I’d recommend it, as I currently think it’s a more leveraged use of time.

(Application form here.)

[Edit Mar 22, 2023: Changed the required commitment from 5 hours to at least 20 hours per week, to reflect the types of projects we’re most excited about at present.]

What content is Open Phil interested in having translated?

In this call for applications, I focus on translating web content — that is, anything available for free on the internet. This includes blog posts, EA Forum and LessWrong posts, podcast episodes, articles, and videos (as long as questions of copyright can be addressed satisfactorily). I’m not focused on books in this post because translating them involves working with the publisher and can’t be done independently, so the work seems less tractable for most people, though also potentially very valuable.

Which pieces of web content are the highest-priority to get translated? Sometimes we can go by ‘track record’ — certain pieces of content have historically gotten a lot of attention and are credited by many respondents to our survey as having been impactful on them or having introduced them to EA/​EA-adjacent ideas. I also have anecdotal data attesting to the impact of other pieces.

Here’s a non-exhaustive list of content I think it’s a priority to get translated, grouped by type but otherwise in no particular order. This is just intended as an example; an item’s exclusion from this list doesn’t mean I think it shouldn’t be translated.

If you apply and we proceed with your application, we’ll likely suggest specific pieces of content we’re interested in getting translated, and make you an offer to work on those.

How does supervising professional translators work? What skills do I need to do it?

The basic plan is that you’d find several professional translators and trial them on a few translation tasks, then select the ones who did a good job and give them more work. In addition to (or rather than) doing translation yourself, you’d be managing one or more translators who aren’t necessarily EAs, and who would be paid for by Open Phil.

It would be your responsibility to find professional translators and reach out to them. Open Phil might be able to recommend possible translators for some languages, but we don’t have special knowledge in the area. We’re happy to give more advice on this process if we move forward with your application.

Conceptually (if everything works out), this is a way to multiply your impact, since one person can review the work of several professionals. I think translations have a lot of community-building value, so in most situations, this seems like a good deal to me (i.e. well worth the monetary expenditure on Open Phil’s part).

Note this is an experiment, and I can imagine it being less efficient than it seems for several reasons — but I think it’s definitely worth trying.

The most important skill in supervising professional translators is being able to recognize high-quality translations. Aside from that, I expect it to involve a small-to-medium amount of logistical work.

How do EAs compare to professional translators on the quality of the products they produce?

In a pilot project at Open Phil, our intern Guille Costa supervised a small number of both non-EA professionals and bilingual EAs in translating content into Portuguese. The finding was that the EAs didn’t have a notable advantage or disadvantage; the quality of the best products produced by EAs and the best products produced by professionals seemed to be about the same, on average, as assessed (blinded) by Guille. This was a small sample assessed by one person, so it doesn’t constitute much evidence.

How will we evaluate applications?

We’ll get back to you within one month of receiving your application. If we move forward with your application, we might reach out to you to schedule a call and/​or to ask you to complete a (paid) work test assessing your translation abilities.

This is an experimental program; we might have to reject applications from people who would do a good job if given guidance or support, simply because we don’t have the time to give that support. Translation into some languages carries more downside risk than translation into others, and this might be a significant input into our evaluation.

What languages are we interested in translations for?

We’re open to funding translation projects in most languages with over five million speakers. The expected impact of a translation varies by language, of course, but we think this is unlikely to be the deciding factor in most applications.

Here’s a non-exhaustive list of languages we’re excited to see (more) translations into:

  • French

  • Portuguese

  • Spanish

  • Russian

  • German

  • Japanese

  • Korean

  • Vietnamese

  • Italian

  • Polish

  • Turkish

  • Persian

  • Arabic

[Edit Mar 21, 2023] Out of this list, we are particularly excited about Japanese, Korean, and Vietnamese translations right now, because we’ve gotten comparatively few applications to do translation work in these languages.

Can translations carry downside risk?

The 2018 post by Ben Todd, “Why Not To Rush To Translate Effective Altruism Into Other Languages” cautions that “most efforts to do translation work should be delayed,” and that instead “efforts to expand EA into other languages should focus on person-to-person outreach to a small number of people with key expertise.”

I agree with the post that, due to political considerations, we should be cautious about doing outreach in (and therefore translation into) some languages. We’ll take this into account when handling applications, and if you have expertise in one of these languages we encourage you to apply regardless.

But I disagree with the post’s implication that EAs should continue to wait to do translations into other languages associated with less dicey political situations, e.g. French, Spanish, or Japanese. I think most languages are substantially less risky.

I share the post’s view that we should make sure translations are high-quality and that translations are best accompanied by “on-the-ground” outreach work — I just don’t think these are reason enough to continue waiting. Ben thinks we’re in a better position to expand into other languages today than when he wrote this post, e.g. see his more recent comment.

Again, apply here if you’re interested in helping with translation work.

Eli Rose23 Aug 2022 23:52 UTC
111 points
37 comments8 min readEA link