Spring ’23 AI Governance, Policy, and Strategy (GPS) Fellowship

MAIA and the Harvard AI Safety Team are excited to announce the Spring 2023 AI Governance, Policy, and Strategy (GPS) Fellowship, a new program aimed at introducing promising Harvard and MIT students to core strategic issues posed by the development of transformative AI systems.

The MAIA GPS fellowship involves two components:

  • A discussion series where participants critically evaluate arguments about risks from the development of advanced AI systems. The series will meet three times during the first two weeks of MIT’s semester. Each meeting will include dinner, reading, and discussion; no reading or other preparation is required outside of the meeting times.

  • A three-day workshop in Washington, DC on President’s Day weekend (February 17-20). At the workshop, fellows will be introduced to core issues in AI governance, policy, and strategy. Fellows will also have the opportunity to meet career professionals working in research, technology policy, and national security. Travel and accommodation will be provided.

By the end of the fellowship, we expect that fellows will be able to contribute to cutting-edge research and policy work in AI GPS topics, including by participating in research organized by HAIST and MAIA.

Apply here by Thursday, February 2 at 11:59pm. The initial application includes both basic information and a short essay; we expect it to take 1-2 hours total. We will interview applicants between Friday, February 3, and Sunday, February 5, and announce decisions on February 6.

  • No technical background is required, though it is a plus. We are mostly looking for curious, truth-seeking people who are clear communicators and have thought deeply about how governments and societies solve problems or why they don’t. Graduate and undergraduate students from all academic disciplines are welcome. People who completed the HAIST Intro Fellowship or the MAIA AI Safety Fundamentals program are especially encouraged to apply.

  • We expect to accept one cohort of up to 8 fellows each at Harvard and MIT. We plan to accept applicants who demonstrate high degrees of interest, thoughtfulness, and reasoning and communications abilities, which adds up to being a pretty high bar.

  • The MIT-based discussion series will meet over dinner for two hours, either on a Monday-Wednesday or Tuesday-Thursday schedule, with exact timing depending on participant availability. We will collect availabilities during interviews.

  • The discussion series will build fellows’ understanding of the best arguments on questions like:

    • Can current techniques in building AI systems lead to human-level generalized intelligence?

    • If so, would this pose an existential risk to humanity?

    • If so, how far in the future should we expect this development to happen?

  • The workshop will include:

    • Reading and discussion sessions on policy topics

    • Q&As and social events with AI GPS professionals

    • A role-playing-based simulation of public and private responses to an AI-related crisis

    • Interactives designed to sharpen participants’ practical problem-solving abilities

    Topics to be covered include:

    • Whether advanced AI systems will tend to produce unipolar or multipolar distributions of power

    • The strategic implications of the rate of AI progress for states, corporations, and interest groups

    • The inputs to AI progress (hardware, algorithms, and data) and the opportunities these present for governance

    • Possibilities for international coordination on governing AI systems