So far the 2 answers you've received are killing people and sending emails.

I don't think either of those are particularly valuable to the society I'd like to see us build.

We're already incredibly dialed in and efficient at killing people. I don't think society at large reaps the benefits if we get even better at it.

Isn't it obvious? Near future vision-language-action models have obvious military potential (see what the Figure company is doing, now imagine it in a combat robot variant). Any superpower that fails to develop combat robots with such AI will not be a superpower for very long. China will develop them soon. If the US does not, the US is a dead superpower walking. EU is unfortunately still sleeping. Well, perhaps France with Mistral has a chance.

Better thank you emails I think. Think how good they'll be on a 10 year timespan

First mover advantage for important AI tools that deliver enormous value to humanity.