I agree with your first 2 bullet points. I agree with the third, with the caveat that doing so leads to a greater risk of error. I'm a consequentialist with utilitarianism as a subset of my values. I think "culpable" refers to how society should treat people, and treating people who fail to save lives as murderers is infeasible and unproductive. I choose TORTURE over SPECKS in the relevant thought experiment, if we stipulate that there are 3^^^3 distinct possible people who can be specked, which in reality there aren't. I want to sign up for cryonics. I like EY's metaethics, and agree that lots of people are crazy.

In addition to that: I think UFA is a risk worth worrying about and working to mitigate and FAI is worth pursuing. I think EY's sequence on words is true, and makes conversations much easier when both people have read it.