Category: AI interpretability

Recent analysis of large language models is reshaping the AI interpretability landscape. In a 26,000‑word essay titled “Neural Networks Are Already Clean,” Nick Cammarata argues that the long‑standing view of AI as an inscrutable black box may be…

Researchers are making progress in mechanistic interpretability, the field that aims to open the black box of large language models. Recent work shows that even at the scale of GPT‑4, analysts can pull out “interpretable features” – human‑readable…