ConceptAI & ML
Harm's law
A measurement of the compute overhead incurred during training when choosing to train smaller models past the Chinchilla optimal point.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Compute overhead1Scaling laws1
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.