avturchin comments on Achieving AI alignment through deliberate uncertainty in multiagent systems