“Scott Wu predicts that AI is rapidly moving towards a point where Reinforcement Learning can train a model to achieve a 100% score on any well-defined benchmark.”