Experiments and essays: reinforcement learning, language models, pragmatics, agents, and one baby name ranker. The cryptographic experiments live under cryptography.
Recent
Unfold your Agents
Most agents should be views. The orchestration problem was solved before most of us were born.
Earnest
A baby name ranker for people who find 400,000 options exhausting.
The Literal Listener
A pragmatic speaker, in Chris Potts' sense, reasons about a literal listener. Train one, and it will learn what the listener likes. That's not the same thing as learning what the listener knows.
Ultimately the Survivors Do Not Prevail
Training multi-agent reinforcement learning in a zombie game taught us something that wasn't really about zombies: the environment is a more powerful programming language than the reward function.
Earlier
Lords Ipsum — GPT-2 Hansard
Fine-tuning GPT-2 on Hansard transcripts to generate text in the style of the House of Lords.
Indirect Programming
Changing Expectations on Machine Program Expressibility.