We are constantly reminded that LLMs are the future despite the real world evidence to the contrary. Look at what happens when LLMs are trained on the output of other LLMs, such as the low quality code flooding the internet. It is all a self-solving problem set in motion.
