I think this is one approach to AI safety and interpretability that could work, but would require labs to slow down to figure out how to extract circuits/algori

I think this is one approach to AI safety and interpretability that could work, but would require labs to slow down to figure out how to extract circuits/algorithms out of trained LLMs rather than deploying the opaque artifact.