The hope is that we will teach the AI as a coach so that it can turn this consensus into a curriculum to teach future AIs so that we can democratically align any language model by training a low rank adapter that makes the new model behave according to the collective will during the deliberation and the rough consensus.