Simplified bio-anchors for upper bounds on AI timelines — LessWrong