AI Governance Policy Fellowship

About CAIAC and This Fellowship

CAIAC's mission is to produce researchers who will address technical and policy questions that reduce risks from advanced artificial intelligence and steer the trajectory of AI development for the better. The Policy Fellowship helps fulfill this mission by offering an introduction on AI safety, while simultaneously preparing rising scholars to think critically about foundational questions and generate new research ideas.

This semester, there are four Fellowship sections (two Technical ML Fellowships and two Policy Fellowships). Each section is led by a facilitator who has tailored the curriculum to fit their strengths. The section I will lead will be treated as a graduate survey course and will focus on generating new research ideas in AI policy, governance, and social impact.

My purpose in focusing on generating new research ideas is to support the mission of CAIAC to produce researchers who will eventually work in AI. Before one works at a major AI lab like OpenAI, Anthropic, or DeepMind, one usually has to have participated in one or several tier one research fellowship programs like MATS, ERA:AI, CBAI, LASR, or Pivotal. These are prestigious research fellowships that have a strong track record of placing fellows into full-time roles at major AI labs. The bar of entry is significantly high to be considered for one of these research fellowship programs, as they are tailored for value-aligned people with a strong research background and record. CAIAC aims to fill this space by offering an initial launching pad that can provide rising scholars with the resources and mentorship to matriculate to a tier one research fellowship and to work at a major AI lab. Several of our alumni have successfully done this and gone on to work at Redwood Research, Constellation, MATS, CHAI, SPAR, and XLab.

What I'm Asking of You

To make this Fellowship as productive as possible, I have two main asks:

1. Please come to the section having read that week's readings. This is a basic responsibility that will be asked of you as you go throughout your academic career and while working at a research lab. The sad alternative of not having read that week's readings is to sit in silence for our allotted time together as everyone tries to catch up and have a conversation that resembles the breadcrumbs that fall off the table after supper.
2. This seminar is designed to generate new research ideas. To do so, the course will push you to understand key ideas in an expansive literature, and you will be asked to submit two short papers that meaningfully extend the ideas in the readings. You may choose to write an empirical or conceptual short paper.

   For an empirical paper, you should outline an original empirical study somehow informed by or related to the week's readings. Your summary should be organized in the following four sections: (1) a research question (1-2 sentences), (2) a description of methods (2-3 paragraphs), (3) your prediction(s) (1-2 sentences), and (4) why the study would be interesting and/or important (2-3 sentences).

   For a conceptual paper, you should advance one or more of the week's readings in some way. Examples include: (a) connecting ideas from two or more papers to develop a more general theory; (b) critiquing ideas in one or more papers and proposing an alternative that you argue is better or more valid; (c) developing a rival theory to challenge the ideas in one or more papers. In your conceptual paper, underline your main thesis statement(s) (1-2 sentences).

Weekly Schedule

Contact

For questions about the fellowship, please reach out to eab2291 -at- columbia.edu