Summary
Brad Carson — former U.S. Congressman and senior Department of Defense official who oversaw the Pentagon’s law of war policy — sits down with Machine Learning Street Talk to push back against the prevailing fatalism in AI discourse. His central argument: AI is not an unstoppable freight train. Societies have successfully constrained powerful technologies before, from nuclear arms control to a global moratorium on human germline editing, and there is no principled reason AI should be exempt from deliberate, democratic governance.
Carson draws on his time running the U.S. Army to discuss autonomous weapons, explaining that his team sent military lawyers to Geneva to negotiate with the Red Cross on possible treaty frameworks well before the topic entered mainstream debate. He also addresses U.S. chip export controls as a concrete lever — not merely a trade tool — for slowing rival nations’ development of advanced AI systems. The conversation ranges into AI-assisted targeting (he cites a 73% probability threshold used to designate Hamas combatants in Gaza as a stark example of algorithmic life-and-death decisions), the dangers of anthropomorphizing language models, and why the American military’s preference for technological solutions over human presence has historically underperformed in counterinsurgency.
For anyone tracking AI policy and governance, this episode offers a rare perspective from someone who has operated at the intersection of legislative, executive, and military power. Carson’s case — that international coordination, not resignation, is the appropriate response to transformative AI — makes it essential viewing for understanding the regulatory debates taking shape in Washington and Geneva.
📺 Source: Machine Learning Street Talk · Published May 31, 2026
🏷️ Format: Interview
