Recursive Criticism and Improvement

It's another example of of emergence, right? Like we used to think of emergence as the thing where you make a bigger version of an old model and all the sudden it has new capabilities that nobody predicted. Yeah, what I found really interesting about this recursive criticism and improvement paper was it's really similar in spirit to that anthropic constitutional AI thing. And it's sort of interesting to see people experimenting more with this idea of using language models to kind of self correct or to correct other language models.