ALTER founder David Manheim argues AI evaluations must be standardized as they become load-bearing for policy and safety

Great initiative to establish common understanding around best practices around conducting and reporting evaluations, as well as challenges to be overcome. V cool to see a delphi process used as part of this.