This spring, an AI chatbot hallucinated that a Chinese ship carried nuclear weapons parts, and US troops and warplanes nearly moved to board it — until a few experienced humans, with minutes to spare, chose to check the machine’s work and found it false. In this segment we look, calmly, at the most important people in that story: not the machine, but the humans who exercised an old and unglamorous instinct — doubt. Why the danger was never the machine’s confidence but our growing temptation to mistake confidence for correctness, why we’re being quietly trained to stop checking, and why the stubborn human habit of the second look — asking whether “sure” is the same as “right” — matters more, not less, the better our machines get.
