Practical AI Alignment and Interpretability Research Group
The Practical AI Alignment and Interpretability Research Group (also known as Pr(Ai)²R Group) develops theoretical frameworks and applied methodologies for uncovering and inducing interpretable algorithms in deep learning models. Founded by Atticus Geiger following his Stanford PhD on causal structures in neural networks, the group conducts original interpretability research, creates open-source educational materials on mechanistic interpretability, and runs mentorship programs. It is funded by Open Philanthropy as part of their work on potential risks from advanced AI.
Funding Details
- Annual Budget
- $368,500
- Monthly Burn Rate
- $30,708
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $737,000
- Fiscal Sponsor
- -
Theory of Change
By developing rigorous mathematical frameworks (particularly causal abstraction) for understanding how neural networks compute, and by training more researchers in mechanistic interpretability methods, the group aims to make it possible to verify AI system behavior, detect misalignment, and build safer AI systems. The open-source course materials and mentorship programs multiply impact by expanding the community of researchers able to perform interpretability work on frontier models.
Grants Received
from Open Philanthropy
Projects
No linked projects.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 12:01 AM UTC
- Created
- Mar 20, 2026, 2:34 AM UTC