Discussion about this post

User's avatar
Jack Shanahan's avatar

Terrific discussion!

Here's how I've been talking lately about the potential for a 'magic bullet' AI in the military: In the extreme, or deus ex machina, scenario, state leaders may believe they possess AI Black Swan military capabilities that offer an insurmountable advantage against adversaries. The promise of a decisive, seemingly effortless victory – a technological decapitation strike, so to speak – could seduce leaders into overestimating AI’s reliability and strategic impact. We’ve already seen how overconfidence in AI can lead to catastrophic outcomes in other domains: high-frequency trading failures in stock markets and flawed predictive policing are just two examples. It’s not difficult to imagine state leaders being swayed by technological determinism and promises of AI-enabled quick wins, with few if any negative consequences or the potential for serious blowback.

The danger here is twofold. First, AI systems are certain to fall short of performing exactly as intended. Second, and perhaps more critically, war is never a one-sided contest. A thinking, adaptive adversary always gets a vote. For every AI-enabled action in warfare, we should expect an AI-enabled counter-action to follow.

Expand full comment
1 more comment...

No posts