Skeptics use it to push back against claims of AI reasoning or intelligence, pointing out that the core training objective is predicting the next token. Others counter that this framing undersells what emerges from that objective at scale, which is a live and unsettled debate rather than a settled fact either way.