Statistics after the loss of innocence: New rigor in the age of AI

This article examines the evolution of statistics in the face of artificial intelligence, pointing to the need to shift from traditional inference to hybrid systems. Drawing on the work of Tianyu Zhan, the author discusses the role of deep neural networks as tools that support, rather than replace, statistical thinking. A key concept is safeguarding—a set of safeguards protecting the integrity of clinical trials from algorithmic errors. The text emphasizes the importance of computational infrastructure and new regulatory guidelines, such as ICH E20, in maintaining methodological rigor. In the age of AI, statistics is becoming an engineering discipline, where managing risk and type I error requires precise decision-making system architecture and quantitative pre-specification.