Looking for feedback on proposed AI health risk scoring framework — LessWrong