I do the human part of Reinforcement Learning with Human Feedback. My list of interests is ever-growing, and this will always be incomplete, but a glimpse of topics that excite me:
- Most things in AI, but particularly LLMs and their uses (like 20% of my read-articles list was SimonW articles)
- Most things in math, with focuses on:
- Combinatorics
- Abstract algebra
- Numerical analysis
- Probability and statistics
- Coding and information theory
- Category and type theories (more of a want to learn than something I've learned yet)
- Game theory
- Card games. I've played Yu-Gi-Oh since I was in short pants and have dabbled in at least all the major TCGs. I also love poker and variants like Omaha Hi-Lo and 7-card Stud.
- Learning something new. I love to figure something out.
I want to make tools that make learning easier for everyone. I'm currently trying to accomplish that through:
- Building Ozymandias, a knowledge base and data processing pipeline, so people like me who want to read literally everything they find can keep track of it, figure out what's unimportant, and take notes for the future. Yes, I know {insert note-taking app} exists.
- Building FixedPoint, a link classifier that may or may not become part of Ozymandias.
- Figuring out automations, mostly in the shell. Recommendations or suggestions welcome.
- Writing sporadically about what I'm doing or learning about.
I develop on an M2 Mac, using kitty terminal, tmux multiplexer and Neovim as my editor. If I'm going to write something, I'm usually going to reach for Python unless I'm trying to learn another language or I need the performance.