Verification Is Not Knowledge: Jessica Hullman on Understanding and Accountability in AI Science

In a guest article for the Cosmos Institute, Jessica Hullman of Northwestern University argues that equating verifiably correct results with scientific knowledge is a misconception, an intuition rooted in the "value-free ideal" that has prevailed since the Cold War. Using Terry Tao's thought experiment on AI cancer therapies and papers on arXiv whose authors struggle to explain their own conclusions as examples, she points out that once verification goes wrong, no one can be held accountable. She argues that resistance to the influx of AI outputs into the scientific record is neither anti-AI nor anti-science.