Dan Hendrycks

Dan Hendrycks is the measurer who became a warning. As an undergraduate at the University of Chicago he co-introduced the GELU activation function that now runs quietly inside an enormous fraction of the models deployed today; as a doctoral student at UC Berkeley he released MMLU, the sprawling fifty-seven-subject examination that became the standard yardstick of a generation of AI systems. Then he looked at the curve those instruments had drawn—climbing toward and past human performance faster than anyone predicted—and asked the question most of the field preferred to defer: if this continues, does humanity keep its hand on the wheel? In 2022 he co-founded the Center for AI Safety, and in May 2023 he orchestrated a 22-word statement signed by the godfathers of deep learning and the chief executives of the leading laboratories, placing the risk of extinction from AI alongside pandemics and nuclear war as a matter of global priority. His answer to that risk is neither doom nor