Joar Skalse comments on Using (Uninterpretable) LLMs to Generate Interpretable AI Code

[Joar Skalse]

Note that this proposal is not about automating interpretability.