zlacker

My take is that journalism provides the pre-vetted training data for the slow AI that is society. (And consequently the fast AI that is LLM which read all the journalism ever recorded)

It’s why it’s important that journalists focus on “this is true”. It’s why science papers are journalism. It’s why it’s important that science papers are published solely on the basis of “is it true” not “does it follow a narrative (#).

This is the huge test for AI/LLM; If you state a thing, can you provide a link for why you know that. At the moment we just have “based on things I have read the thing is statistically most likely to be said (hence true).”

We really need “the truth”

(#) I am thinking here of things like archeology in Nazi Germany which switched heavily to publishing nonsense about Aryan people in different parts of world earlier than anyone else. Not some comment on diversity and inclusion.

>>lifeis+(OP)
It's important to note that journalism tends not to operate on a truth standard, but on an attribution standard, and holds itself to a "balance" which requires finding one liar to match every true statement, like a shitty logic puzzle.

And of course stuff on newspaper websites that's labelled "opinion" is under no standards at all.

> Not some comment on diversity and inclusion.

"Archaeology should not be Nazi" is very much a DEI statement! If a very basic one.

>>lifeis+(OP)
My old journalism professor, Bob Woodword ("Not that one," he'd chuckle) would constantly chant "Accuracy, accuracy, accuracy" as he taught. I think he chose those words carefully because he recognized that while all news is going to have a slant, at least you can be accurate in presenting it. I don't know if LLMs will ever be able to see the facts through the slant.