Welcome to my digital garden!

My name is Ed. I am a researcher in alignment and quantitative finance, currently pursuing a PhD in Mathematics at the National Academy of Sciences.

In this garden, I share my daily journey in tackling one of the most important problems of our time, along with occasional side stories.

If you are looking for a researcher to collaborate with or would like to discuss ideas, feel free to reach out to me here. I have a wide range of interests within the field of alignment.

Projects:

• Cyclic ablation paper: — Empirical proof that iterative ablation against deception is ineffective; +1 to the manifold hypothesis.
• Foundation white-paper:— Addresses open/closed AI governance and proposes an economic incentive framework for alignment. Currently applying game theory to transform this into a formal paper.
• Mechanistic Unlearning: Developing a technique leveraging geometric similarity via Gromov-Hausdorff distance methods to detect and suppress hazardous subspaces.
• Defining agentic preferences — As part of the SPAR program, I formalize a family of agency metrics and provide convergence estimations.
• Self-led agent behavior: Examining the behavior of self-led agents (also part of the SPAR program).