Federal judge denies Meta's motion to dismiss Strike 3 Holdings' lawsuit over torrented adult videos used for AI training

latest update from the pr0n v meta lawsuit: meta's motion to dismiss is DENIED!

this would be an interesting precedent to set if distribution of copyrighted material eventually gets enforced (expanding to also include outputs from models & decimating all fair-use coverage)

though like gdpr, at worst it might just end up being a trivial fine

case update: https://www.courtlistener.com/docket/70899478/61/strike-3-holdings-llc-v-meta-platforms-inc/

404 coverage: https://www.404media.co/judge-rules-blacked-com-can-sue-meta-for-scraping-its-porn/

@GaryMarcus I really can't wrap my head around this one. Why would Meta want to train its AI on porn? Did Zuck need a training program on what a dick is for, rather than just being one??

@GaryMarcus Oh man.

Consider: they presumably ripped zillions of pornsites. So, not only is all the porn in there, but all the CSAM, revenge porn, creepshots etc that wasn't picked up/flagged is IN THE TRAINING SET.

Meta may have made itself the world's biggest user/maker of illegal porn.

@GaryMarcus Because they would have just scraped it willy-nilly. We know that IP means nothing to these people. There's little to suggest they would have cared to filter it (or been able to filter it) particularly stringently.

@GaryMarcus I look forward to discovery just because a judge will have to read the titles out live in court.

@GaryMarcus the articles all had the exact same wording too lol

they really just submitted whole porn studio archives

@GaryMarcus Amusingly, FAFO fits this situation perfectly @Meta . #generativeAI #aigenerated #artificialintelligence #AI #slop #facebook

@GaryMarcus Begs the question of what they were training the AI model for with porn?

@justexcel @GaryMarcus Good zinger, but you've hit the nail on the head. Why in God's name did they scrape pornsites? What did they hope to achieve beyond being served for a big class-action?

@justexcel @GaryMarcus They (and I mean all corpos invested in development of generative A.I. technology) need massive data. As in as much data they can get their hands on, regardless of the source. It's why no LLM can work with just public domain content (volume is too small).

And data degrades, too.

@CadmanK91706 @GaryMarcus why should they? how else would you research the internet then to scrape it?

@suchenzang I love the detail that they seeded everything they grabbed. Thank you for practicing good torrent etiquette Zucc