This paper defends anti-AI-natalism, the view that creating sentient AI is impermissible due to the risks imposed on the AI itself. A powerful objection to this position holds that it entails an acceptance of anti-natalism, the view that human procreation is impermissible. Against this suggestion, I will argue that anti-AI-natalism can, and should, be accepted alongside a rejection of anti-natalism. To make this case, I draw on Seana Shiffrin’s account of anti-natalism, arguing that while it fails in its intended context of procreation, it succeeds when applied to the creation of sentient AI. The morally significant difference is that while a procreator can reasonably expect their child’s subsequent and undeluded endorsement of their creation, we cannot reasonably maintain the same expectation in the context of creating sentient AI. Therefore, while it is (ordinarily) permissible to procreate, it is impermissible to create sentient AI.
L. Lang (Thu,) studied this question.