I'm not sure that dismissing government reform is necessarily the right thing to do, even if AI-risk is the larger problem. The timelines for the good the solutions do may be different - even if you have to save the world fifty years from now from UFAI, there's still a world of good you can do by helping people with a better government system in the meantime.

Also, getting a government that better serves its constituents' values could be relevant progress towards getting a computer program that serves well the programmers' values.

You've probably thought through these exact points, but glossed over them in the summary. It's still worth mentioning, even though you've already bought the educational options needed to tackle the AI problem.