Roko comments on What actual bad outcome has “ethics-based” RLHF AI Alignment already prevented?

[Roko]

alignment has always been about doing what the user/​operator wants

Well it has often been about not doing what the user wants, actually.