Joseph Miller comments on Against Yudkowsky’s evolution analogy for AI x-risk [unfinished]

[Joseph Miller]

See No convincing evidence for gradient descent in activation space