LLM is a lossy compression of the internet. We can provide it in a form that is directly executable on 80s computers though gpt4 tries to convince me that it is

LLM is a lossy compression of the internet. We can provide it in a form that is directly executable on 80s computers though gpt4 tries to convince me that it is practically impossible the reduced model would be much weaker (somebody doesn't want to be sent to 80s ;)