The key "AI" contribution, after you dig down through all the media misunderstanding, appears to be this: [NYT]> The time it took for the camera images to reac

The key "AI" contribution, after you dig down through all the media misunderstanding, appears to be this: [NYT]> The time it took for the camera images to reach the sniper and for the sniper’s response to reach the machine gun, not including his reaction time, was estimated to be 1.6 seconds, enough of a lag for the best-aimed shot to go astray. So aiming and firing was initiated by a human, but hardware and software attempted to mask the time delay in doing so against a moving vehicle. Which seems plausible: optical flow + basic vehicle vector and velocity calculations. There's also some stuff in the various articles about facial recognition, but that sounds a bit far fetched and unreliable. If you have a person in the loop with enough resolution to accurately aim, then why are you running facial recognition to discriminate targets?