The US rejected the idea of a treaty to ban autonomous robots with lethal arms.

I can imagine US officials thinking about the current US advantage in AI and concluding that unrestricted deployment of autonomous weapons would give the US an advantage in this decade, so (they might conclude) it is better to reject the treaty.

On the other hand, if 15 years from now China is in the lead in AI, having this treaty in place already would benefit the US against China.

Those short-term considerations are the wrong way to look at this question, because it is bigger than the narrow questions of short-term advantage. In the long term, this treaty will prevent a danger to all humans.