Disappointing.
The first half of strong with a history of autonomous weapons, and a great taxonomy of the different types of autonomy, hinting at how the lines are both more blurry and more historic than I expected.
With a strong setup, I was hoping for a cogent analysis of how the potential for increased autonomy via ML could change the economics or reality of warfare, instead the author gives a regurgitation of sci-fi tropes and overwrought philosophizing about the morality of autonomous weapons. Scharre convinced me that peacetime pontificating was not the most relevant line of inquiry after describing US moral reluctance to engage in submarine warfare, completely scrapped mere hours after Pearl Harbor. In government and private industry, I’ve watched years of careful thought scrapped when faced with the reality of defeat.
For an analysis of the ways in which ML can go wrong, Weapons of Math Destruction was better, and for speculation of what autonomy could affect on the battlefield, I was much more impressed by speculative fiction such as Daemon, or The Last Good Soldier.