The U.S. Should Rely on Performance, Not Explanation, When Evaluating AI (op-ed)