Technical Governance Researcher

Technical Governance Researcher ABOUT MIRI The Machine Intelligence Research Institute https://intelligence.org/ (MIRI) is a nonprofit based in Berkeley, California, focused on reducing existential risks from the transition to smarter-than-human AI. We've historically been very focused on technical alignment research. Since summer 2023, we have shifted our focus towards communication and AI governance. See our strategy update post https://intelligence.org/2024/01/04/miri-2024-mission-and-strategy-update/ for more details. ABOUT THE TECHNICAL GOVERNANCE TEAM We are looking to build a dynamic and versatile team that can quickly produce a large range of research outputs for the technical governance space. Please feel free to fill out this form, or contact us at techgov@intelligence.org. We focus on researching and designing technical aspects of regulations and policy that could lead to safe AI. The team works on: - Inputs into regulations, requests for comments by policy bodies (e.g. NIST/US AISI, EU, UN) - Technical research to improve international coordination - Limitations of current AI safety proposals and policies - Communicating with and consulting for policymakers and governance organizations Our previous publications are available on our website https://techgov.intelligence.org/research if you would like to read them. See our research agenda https://arxiv.org/abs/2505.04592 for the kinds of future projects we are excited about. ABOUT THE ROLE In this role, you would have the chance to work on all of the above areas. The work will be a mixture of researching, writing (for internal and external use), and engaging with collaborators and policymakers. Most of the day-to-day work is a combination of reading, writing, and meetings. Some example activities could include: - Threat modeling, working out how AI systems could cause large-scale harm, and hopefully what actions could be taken to prevent this - Responding to a US government agency’s Request for Comment - Learning about risk management practices in other industries, and applying these to AI - Designing and implementing evaluations of AI models, for example to demonstrate failure modes with current policy - Preparing and presenting informative briefings to policymakers, such as explaining the basics and current state of AI evaluations - Reading a government or AI developer’s AI policy document, and writing a report on its limitations - Designing new AI policies and standards which address the limitations of current approaches - In the above work, maintain particular focus on what is needed for solutions to scale to smarter-than-human intelligence and conduct research on which new challenges may emerge at that stage WHO WE’RE LOOKING FOR There are no formal degree requirements to work on the team, however we are especially excited about applicants who have a strong background in AI Safety and have particular previous experience or familiarity working in (or as) one or more of: - Compute governance. Technical knowledge of AI hardware / chips manufacturing and related governance proposals. - Policy (including AI policy). Experience here could involve writing legislation or white papers, engaging with policy makers or other research in AI policy and governance - Strong AI Safety generalist. For example, you have produced good AI safety research and have a good overview-level understanding of empirical, theory and conceptual approaches, or otherwise have a demonstrated ability to think clearly and carefully about AI safety. - Bonus: Research or engineering focused on frontier AI models or the AI tech stack. The role may involve creating or running model evaluations, benchmarking AI hardware, conducting scaling law experiments, and other empirical work. We are also excited about candidates who are particularly strong in the following areas: - Agency – You get things done without someone constantly looking over your shoulder. You notice problems and ... (truncated, view full listing at source)