[richardjennings]

Is it possible that LLMs could challenge Data Compression Information theory ? Reading this made me wonder how much can be inferred via understanding and thus removed from the minimal necessary representation.