The Ex-Pentagon Chief Sounding the Alarm on AI Weapons — Brad Carson
Machine Learning Street Talk · 1:20:52 · 1 months ago
AI is not an inevitable or unstoppable force; it is a technology that society can and must shape through active governance, especially to prevent the erosion of accountability in warfare and to address the growing loss of public trust.
- Control over development — The belief that AI is impossible to stop is a dangerous myth; because Western nations control the essential chip supply chain and have successfully restricted technologies like recombinant DNA before, we have the power to regulate its trajectory .
- Erosion of accountability — Applying neural networks to warfare replaces clear, binary decision-making with opaque, probabilistic calculations, creating an accountability vacuum where soldiers cannot be held responsible for the machine’s flawed output .
- Product liability — Rather than treating AI like a human or a new legal person, regulators should classify it as a manufactured product, holding companies liable for design flaws that enable misuse, just as they do with other consumer goods .
- Transparency mandates — Current AI policy relies too heavily on informal influence networks; mandatory testing and independent verification are necessary to ensure safety and prevent companies from operating with zero oversight .
- Public perception — AI is currently viewed by many as an elite-driven project designed to disrupt their lives and threaten their jobs; for the industry to endure, it must rebuild trust by proving it creates broad, positive value for the public .
- Diplomatic engagement — Nations should prioritize "track two" diplomacy—informal discussions between former officials and experts—to understand the intentions of adversaries, similar to how Cold War-era nations managed nuclear tensions .