“A Complete Overview of Our AI Safety Talent Pipelines (+ Interactive Visual!)” by Roman Ross

Summary Making AI safe and good requires many people to be ready to address crucial problems. However, our current “pipelines” for finding, training, and accelerating talented individuals who could work on these problems are “leaky”, meaning there are many ways they can be improved. This post shows my full overview of the AI safety talent pipeline: where people come from, where they go to learn more, and why this is impactful. I also describe some “leaks” in the pipeline: stages that are functioning below their best level, and might be improved with some effort. As an alternative to reading this boring EA Forum post, you can instead get all of this information from this interactive visual I vibecoded. Seriously, the visual is my recommended way of reading this post. Introduction A few quick notes on how to read this post: I originally created this document as a “personal checklist” of fieldbuilding tasks that seem good to complete,[1] and I created the associated website because I thought it would be an efficient way to store that information. I am publishing this because I’ve learned a lot about fieldbuilding needs along the way, and I thought some people would [...] ---Outline:(00:12) Summary(01:12) Introduction(01:15) A few quick notes on how to read this post:(02:31) Categories of People(03:00) The people:(03:24) First Discovering AI Safety(03:57) Funnels (How People Initially Learn about AI Safety)(04:02) Standard Media Outreach(05:10) AI Safety and Adjacent Internet Cultures(05:46) Adjacent Activist Groups and Protests(07:09) University Groups(07:50) Fellowships(09:06) Recommendations from Friends(09:48) Events(10:13) Who gets lost?(11:12) Here are some projects that might be worth trying (in addition to fixing leaks):(12:56) Really Starting to Care about AI Safety(13:30) Events:(14:48) Other Communities:(15:10) Online Resources:(15:55) Who gets lost?(16:15) Another question: How can we make the process faster?(17:20) Here are some projects that might be worth trying (in addition to fixing leaks):(17:47) Upskilling(20:21) Career Transition Materials(20:47) What are we missing?(21:41) Here are some projects that might be worth trying (in addition to fixing leaks):(22:38) Job Decisions(23:18) Job Decision Resources:(24:56) Here are some projects that might be worth trying (in addition to fixing leaks):(25:13) Theories of Change(25:16) Technical Researchers(26:03) Policy and Governance Workers(26:32) Org Scalers(27:30) Operations People(28:20) Other Generalists(29:01) Communications(30:02) Additional Notes(31:07) General Concerns About This Whole Framing:(33:07) Ending Notes The original text contained 3 footnotes which were omitted from this narration. --- First published: July 17th, 2026 Source: https://forum.effectivealtruism.org/posts/gYJJKea4DNs89iEbb/a-complete-overview-of-our-ai-safety-talent-pipelines --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.