normative
Interpretability As X-Ray Not Intervention
Mechanistic interpretability is valuable because it functions like an X-ray or assessment of a model rather than a modification or intervention; alignment methods form an 'extended training set' that can be gamed, while interpretability provides an 'extended test set' that the model isn't optimizing against — but you must never train for interpretability or look at the test set too often, or you destroy that advantage.
normativepending
Speaker
Dario Amodei (Anthropic CEO)Evidence Quote
“if you look at the X-ray too many times, you can interfere. We should worry about that, but that's a much weaker process, it's not automated optimization.”
Source
Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough— Dwarkesh Patel PodcastCreated: 6/13/2026, 3:25:28 AM
Your notes
Loading notes...