I wouldn't count on computer expertise being infinitely available even after AI is invented. They won't be infinitely intelligent; we will still need to give the programming AIs some sort of spec, which will need to be written by people who understand how the AI thinks. If nothing else, there will need to be people who understand it well enough to make sure it's actually serving peoples' interests.

More likely, there would be a gradual transfer of control to increasingly sophisticated AIs, paralleling our current/past move to higher-level languages.