Chris Painter
My current vision for a good AI safety outcome looks more like "All of the necessary external testing organizations and institutions exist, with the right combination of assessment, oversight, and enforcement powers" and much less like "A well-understood set of clear safety standards exists that is easy to run and export". This means that the US might have some AI safety/governance infrastructure available to it that China doesn't have. If China wakes up late to the importance of alignment and external technically sophisticated institutions, then it will be going through the same capabilities inflection points (on a delay) that the US is without the same contingent institutions.