Opening scene :
Trust machines … Speed better than control
AI compresses Boyd’s OODA loop beyond human cognition. AI-assisted targeting identifies, classifies, calculates, and guides in seconds. The human-on-the-loop vs. human-in-the-loop distinction is the central policy question. In high-density environments with hundreds or thousands of simultaneous drones, individual human approval is physically impossible. A 500-drone swarm requires decentralized autonomy or a command bottleneck negating speed advantage.
The logical endpoint: AI making majority tactical engagement decisions, with human oversight focused on strategic guidance and exception handling. Ukraine’s stated goal is autonomous operations with human involvement only at the final engagement decision. This complies with emerging international guidelines while maintaining AI’s speed advantage.
Risks are not theoretical. AI operating at machine speed on both sides could produce escalation spirals outpacing human understanding. Misidentification, friendly fire, and unintended civilian engagement all increase with scale and speed. The December 2022 friendly-fire incident where USS Gettysburg shot down an F/A-18 Super Hornet responding to Houthi drones illustrates how human-machine coordination failures occur even without AI autonomy. Full autonomy magnifies these risks by orders of magnitude.