Lorec comments on Is theory good or bad for AI safety?