Commentary: The problem with pausing data centers

A bill has been introduced in Congress by Senators Bernie Sanders and Representative Alexandria Ocasio-Cortez, aiming to halt the construction of data centers essential for advanced AI. The proponents argue that the rapid pace of AI development poses significant risks, necessitating a pause to ensure safety and allow governance to catch up. However, a commentary published in The Daily Gazette critiques this proposed policy, asserting that pausing data center construction would effectively pause AI development itself. The authors, Chris Koopman of the Abundance Institute and Keving Frazier of the University of Texas School of Law, argue that technological progress is incremental and iterative, and that stopping development would prevent the discovery of constraints and the building of institutional knowledge necessary for effective governance and improved safety. They cite examples of past failed predictions of AI catastrophes, noting that systems like GPT-4 have evolved through deployment, leading to improvements like reduced hallucination rates. The commentary highlights that despite previous calls for a pause on AI development, no feasible governance regime was ever proposed. It suggests that market forces, institutional adaptation, and practical deployment experience have instead led to improvements in safety practices. The authors conclude that the underlying logic behind the proposed data center pause mirrors earlier, mistaken arguments for halting AI development, and that such interventions misunderstand how technology progresses.