Writing
Notes on training a 27B vision-language model on long-context examples.
Reflections on disappearing choices, AI and the experience of work, institutions under pressure, and different experiences of the future.
Lessons learned pushing an LLM-assisted project end to end — what works, what doesn't, and the research behind why LLMs behave the way they do.
Notes from NeurIPS 2019 where I presented the Pommerman work. Includes summaries of talks on RL, NLP, and an informal discussion with Sutton, Silver, Littman at the RL social.
A curated list of blogs and lecture series that I found valuable for learning ML and RL — CMU, Berkeley, Stanford, DeepMind, and more.
Notes from EMNLP 2018 in Brussels — one of the largest NLP conferences at the time. Trends, best papers, and observations on the ELMo revolution.