This brings me back 20 years ago when I first saw markov chains on an IRC bot that would reply to you as well. I was so amazed that even very rarely it could generate valid-ish combinations, of course we are the ones that would interpret it or have to twist it in so many different ways to "make sense of it". But little by little the nonesense wore out and the bot was just understood as a garbage generator, that you would read now and then when you were bored and had really nothing to do but to talk to a bot and see if any of the strings made valid sense, just for fun see when it would generate something that, to one human being, could be a valid point in a big context, literally throwing words in the air and trying to make sense of the resulting sentence as some sort of philosophical game. At least back then it was just 10 or 15 words per generated garbage batch.

The other thing this reminds me of is the book "Children of Memory", funnily enough I was listening to it when the chatgpt craze started, there's two characters that crack me up, two corvids, no spoilers, but they live in pairs and have access to quotes from earth bibliography and live in a distant future in an alien planet, one would quote things such as "my kingdom for a horse", and the other would reply "but you don't know what it means! what even is a horse... We lost that data to make space for something more relevant to our times" and they would hold this discussions about meaning, they really cracked me up and to me chatgpt is one of the corvids, I hope the other corvids arrive soon and we can get some popcorn at the level of nonesense feedbackless garbage discussions, sort of like that beautiful tale of the seven blind men that want to learn what an elephant is like, each touches a different parts of the elephant and hold discussions about how their contextual interaction is the one truth, one talks about the tail, another the husks, another the trump, another the leg and so on...