Hi! I’m Alana.
My goal is to make the future awesome. I currently do AI safety research. I’m on a leave of absence from Stanford University, where I studied Computer Science.
In the past, I have:
- Built systems to catch bad actors for WhatsApp
- Conducted pre-release safety testing on GPT-4 for ARC Evals (now METR)
- Studied worst-case risks from AI at the Center on Long-Term Risk
- Conducted research on the AI alignment problem with Vivek Hebbar via the MATS program
These days, I think about:
- How to test, monitor, secure, and contain powerful future AI systems
- How to build infrastructure and tooling that speed up researchers
- The inner lives of transformers
You might want to…
- Check out alana-utils, my opinionated Python library for LLM-heavy workflows.
- Check out eval-awareness, my ongoing research into LLMs’ ability to discriminate between evaluation and deployment environments.
- Keep up with my every move on GitHub.
- Say hi! (My email address is hi @ my-second-level-domain (dot) my-top-level-domain). I’d love to hear about what you’ve been working on and/or thinking about.