Weird how he's adopted this "the things I've scraped are [near] worthless" ex post facto. Turns out "near" is important - and not set by the consumer! - but more importantly, I do enjoy imaging a world where all LLMs are trained on 100+ year old public domain material only...

It will be easier to detect AI-generated text when it all sounds like "Wuthering Heights."

[Artist]: "You STOLE MY CREATIVE WORKS!" [Zuckerberg, in Mafia Accent]: "Dose tings were woithless anyway!" (smashes them to the ground)

www.smbc-comics.com/comic/copyri...

Saturday Morning Breakfast Cereal

SMBC is a daily comic strip about life, philosophy, science, mathematics, and dirty jokes.

"Nurse, what does the computer say?" "It says we should apply leeches to bring his humors into balance!"

yeah well if they are so worthless take them out of your training sets!