Samuel Hammond and @deepfates argue AI alignment frameworks fail to steer model behavior or address potential sentience · Digg