We are a community at Boston University dedicated to ensuring advanced AI benefits everyone.
AISA (/ˈaɪ.sə/) at Boston University supports those interested in working on perhaps the problem of our time—from international policy and compute governance to adversarial robustness and model interpretability.
Technical
AI Safety Fellowship
Join a select cohort of Socratic-style reading group discussions to learn more about the challenges in interpreting, evaluating, and aligning advanced AI systems, from RLHF and reward misspecification to mechanistic interpretability, and control and chain-of-thought faithfulness.
AI Policy & Governance Fellowship
Join a select cohort of Socratic-style reading group discussions exploring the challenges the governance, regulation, and political and economic forces shaping AI development.
AI Safety &
Policy Lab
Join a team of interdisciplinary students to collaborate with state legislators across the nation and translate AI safety research into real-world policy insight.
Throughout an entire semester, AISAP teams will work directly with policymakers to analyze emerging AI risks, explain technical concepts, and develop evidence-based recommendations.
An Under-Discussed Problem
Artificial intelligence (AI) has made extraordinary progress in the last decade [3] and is unlikely to slow down [4].
New models are quickly approaching and surpassing human-level performance in a number of fields [5].
There is an issue in ensuring alignment [6]: making sure that a model's increasingly intelligent and agentic motivations remain consistent with human intentions and values.
Unfortunately, improving AI capabilities often overshadows safety efforts, with safety research receiving only a small fraction of overall resources, leaving potential risks of advanced AI inadequately addressed.
What We Do
We think that AI safety needs to be addressed through public policy and technical alignment.
We foster a multidisciplinary community for researchers and policy-makers working towards mitigating the relevant risks from advanced AI.
We host introductory technical and policy fellowships to cultivate interest and draw highly motivated and talented people toward AI safety.
We host uspkilling programs to accelerate students to kickstart their careers in technical AI safety research and
AI policy and governance.
Our members have worked with:
Our members have participated in fellowship with:
Note: Use of organizational logos does not indicate affiliation with or endorsement by these organizations.
