Mission would be to create the first general AI and use it for individual empowerment — i.e., the distributed version of the future that seems the safest. More generally, safety should be a first-class requirement.
I think we'd ideally start with a group of 7-10 engineers/researchers, and that we'd want to grow it with great people over time. Initial committee would be: you, me, Greg Brockman. Research leader Ilya Sutskever (the best in the world for our purposes I think; would have to convince him to leave Google).
Governance: I think governance should be a 501(c)(3) with a board. You'd be the chairman, I think. The technology that comes out of this should belong to the world, broadly construed.
We could structure compensation pretty much like a startup — meaningful equity in the form of YC-like profit-sharing units in any commercial entity that gets spun out — but a credible commitment that the technology is going to be widely shared.
Initial funding from YC Research; we'd raise more from outside.
Sam