Artificial Intelligence · AI Alignment & Safety
A topic within AI Alignment & Safety, itself one of 11 topics in that field and part of Artificial Intelligence.
2 works
Olah et al.
Argues neural networks can be understood the way biologists understand organisms: individual neurons and the weighted connections (circuits) between them recur across models and implement identifiable algorithms.
≈9,000 wordslink checked 17 Sept 2026Elhage et al.
Reverse-engineers small attention-only transformers into a linear algebra of composable circuits, and uses it to identify induction heads as the mechanism behind in-context learning.
≈20,000 wordslink checked 17 Sept 2026