4. Safety research. As AI capabilities grow, alignment and safety research should keep up with or ideally lead those capabilities. I liked Sam Bowman's summary of this a lot

→ View original post on X — @_jasonwei