Jet-based hovering vehicles like this and the Flyboard Air suffer from a pair of difficulties: inherent instability and the inability of their turbines to throttle up and down fast enough for basic computer control to keep them stable. Such vehicles are today manual-fly only and require a high degree of skill and talent to operate safely. I wonder if machine learning could be applied predictively to achieve a degree of stability beyond normal "tilting left, apply thrust left" programming.