Let me expand: Tesla lays out how it built the AI system to provide full self driving. It occured to me that some thibgs also apply to AWE in the Yoyo form at least; humans have trouble flying the thing due to information overload and processing and response time. Musk touches these issues in the Q&A session. It seems clear to me that for AWE, keeping the kite airborne is a huge and real issue. For instance, a person cant deduce the apparent velocity at the kite, the wind speed, the angle og attack at all. On the other hand, an AI pilot would have access to all of these and more values. And be able to act in microseconds. And the world of an AWE kite is initially very simple: just the ground, the winch, the tether, the wind and the kite.
I think if someone put just a little effort into an AI pilot, these kites would fly themselves to a larger extent, thus making it simpler to keep it airborne and gather flight hours.
It could also solve the issue when a gust causes the kite out of the planned path, and no preprogrammed routine exists (ie freestyling). The top priority would be never hit the ground, then keep flying speed over a minimum, tether tension below a maximum etc etc.
In the end figuring out where to fly is probably a lot simpler than parking a car at an airport with heavy traffic