AlphaGo's Move 37 and Its Implications for AI-Supported Military Decision-Making
Level 5 - mechanism / opinion, no new human data
Level 5 by design analogy; conceptual analysis and theory with no empirical human data.
OpenAlex W4394686793 · doi:10.1201/9781003410379-15
What was done
The author conducted a conceptual analysis using AlphaGo's "Move 37" from its match against Lee Sedol to define the concept of "unpredictable brilliance" in artificial intelligence. The chapter analyzes how this property impacts military decision-support systems and the governance structures that allocate accountability in combat.
What was found
The abstract reports no empirical or numerical findings. The author argues that managing the risks of unpredictable brilliance—including risks to blue force safety and campaign objectives—redistributes responsibility for combat performance away from military commanders and toward the institutions that design, build, authorize, and regulate AI systems. This structural redistribution is argued to occur similarly for both human-in-the-loop and human-out-of-the-loop systems.
Why it matters
This paper identifies a critical challenge in military AI governance, proposing that simply keeping a human in the loop does not preserve traditional commander accountability when systems exhibit non-intuitive, advanced reasoning.
Limits
This is a purely theoretical chapter based on an analogy to a game of Go with no empirical, experimental, or real-world military operational data. No quantitative performance or safety metrics are evaluated.
Cited by
- supports In game three of the five-game match between AlphaGo and Lee Sedol, AlphaGo played move 37, a move Go masters had not previously conceived.