As an AI learning problem this is probably interesting.<p>But the successive convexification algorithms that they really use for landing are provably correct and will always converge to the right answer, which is not something you can ever say about AI training.