[dontupvoteme]

>There are relatively few places on the internet where that is true

Imageboards.

DailyMail.

Slashdot.

Even a somethingawful dump would have been superior.

[mike_hearn]

Slashdot doesn't have the volume. Don't know about image boards but are they threaded and do they cover as many topics?

The Daily Mail (the newspaper) has been used for training LLMs in the past, yes. I don't know if it still is.

[red-iron-pine]

imageboards sure do -- poorly.

listen, some of the niche corners of that world aren't so bad, but it ain't the place to be training AI to do something, unless that something is a hate crime

[dontupvoteme]

So filter the content before you use it. Clearly openAI did the bare minimum on this front.