The trivialization doesn't negate the point though, and LLMs aren't intelligence.

The AI consumed all of that content and I would bet that not a single of the people who created the content were compensated, but the AI strictly on those people to produce anything coherent.

I would argue that yes, generative artificial stupidity doesn't meet the minimum bar of original thought necessary to create a standard copyrightable work unless every input has consent to be used, and laundering content through multiple generations of an LLM or through multiple distinct LLMs should not impact the need for consent.

Without full consent, it's just a massive loophole for those with money to exploit the hard work of the masses who generated all of the actual content.