Focus Posts
4 posts. Short notes built around one claim I want to make precisely.
-
Using AI kind of involves interacting with other people
A focused note on generative AI use as mediated interaction with upstream human contributors.
-
N-gram search as posterior updating for data attribution
A short argument that model outputs and training-data priors can improve rough approximations of data attribution.
-
Augmentation is a data flow problem
A short argument that "augment, do not replace" is about data control.
-
AI progress as quasi-public good production
A short argument that AI systems pool diffuse human work into privately governed, public-good-like model weights.