Disenchantment with the Post-AI Internet(lukesmith.xyz) |
Disenchantment with the Post-AI Internet(lukesmith.xyz) |
> The AI hoax is the new COVID hoax.
Does the guy imply that he does not believe in a pandemic?
His words about wikipedia and enforced thinking strike me as very absolute, as far as I know, depending on language, the wikipedia editors and admins sometimes have peculiar style choices and stright-ish guidelines, buts its a far cry from idological enforcement.
On the other hand, his point about all the LLMs basically having taken a snapshot of 2023-ish Internet think and soldifying it seems somewhat valid. I remember some reports about a hiring AI-based help that prefered non-colored non-woman humans because that is where the history had its bias.
And as the author I feel kinda sad when I see an AGENT.md or CLAUDE.md in a repo..but reality is rarely that black&white.
While that may be true, it doesn't happen _that_ often, unless you spend most of your day googling random phone numbers...
This is the only sane approach.
I used to like Luke's style, but I think he's spent a bit too much time in the woods. It's a pretty weak comparison to call LLMs a religious experiment.
There aren't too many interesting ideas in this piece, just rambling and complaining. Oh well.
The neat thing about Wikipedia is that you can, actually, edit it. And pages do change over time to reflect new sources and new consensus, after discussion and, sometimes, debate. Wikipedia also strongly grounds its content in citations and secondary sources, while AI slop generally does not.