phasewalk.xyz

Applied Math Undergrad at UC Berkeley · AI Safety and Alignment Research

I believe AI Safety, Security, and Alignment are among the most important problems of my lifetime. I am concerned about our trajectory, where capabilities are far outpacing our capacity for building safe and aligned systems. I want to put my energy into narrowing that gap. I’m drawn to interpretability, evaluations, monitoring and control, red-teaming, character training, and policy research.

I'm exploring some of these questions in a Technical AI Safety class with BASIS, where we read and present research papers, run experiments, and do tabletop exercises exploring AI risk scenarios. Recently, on my own, I’ve been experimenting with characters by training a model on Cayde-6's in-game dialogue in Destiny (blog post coming soon ...!)

Before starting at UC Berkeley, I got my A.S. in Computer Science at Diablo Valley College (DVC). I worked on distributed systems at Nethermind, and I contribute to open-source projects that I care about when I can: like Firefox, Penumbra, Hermes, and Namada.

Some of my other interests are, if I were to cast a wide net: Computing, Cybersecurity and Privacy Tech, Free and Open-Source Software, the @ protocol, VR/XR, HCI and Cognitive Science, Blogs, Skiing, and Electric Guitar.

I like moving between fields and between theory and building things. I'm looking for people to learn from and work with on making the future go well.