Machine Learning - Alpha Dogfights

I was fascinated by the alpha dogfight example that the AI was victorious over the best fighter pilots, but also that they thought of and they executed maneuvers that humans couldn't execute. Why does AI sometimes get things spectacularly wrong? And can you give some examples? I think there are a lot of places where AI does the unexpected. It takes human judgment to understand which of these quote unquote crazy ideas are crazy in a good way. They're highly creative. They're going to push understanding and discovery in new ways...which should just not see the light of day.