Problem is that so much current content is already AI generated that its becoming difficult to train LLM models using the Internet. Going forward, either we find a way to train AIs on AI content without driving it insane, or we keep dipping back into the only deep well of guaranteed human content way into the future.
It will be preserved and replaced within 70 years. I mean, imagine if the Internet was around in the 50s, and we had an AI trained from it.
It would be a novel curiosity, but not the one we go to for up to date information.
So “our” LLM that we’re encoded on will too be updated and replaced.
Not much of an afterlife. Language might not even be the same in 50 years, no cap skibidi.
Problem is that so much current content is already AI generated that its becoming difficult to train LLM models using the Internet. Going forward, either we find a way to train AIs on AI content without driving it insane, or we keep dipping back into the only deep well of guaranteed human content way into the future.