Discussion
World's smallest violin needed, again:
"OpenAI Loses Key Discovery Battle as It Cedes Ground to Authors in AI Lawsuits" —if found guilty of 'wilful" infringement OpenAI could be on the hook for up to $150,000 per work. (And circumstances don't look great for them—they appear to have destroyed lists of the copyrighted works they'd trained GPT on.)
@cstross $150,000 per work?!
Okay, OpenAI, there are at least 20 pages of my work in your database, and I'm too lazy to do the math at this point. I'll accept if you're too lazy to do the math and decide to round up to $2 million. Just use the IBAN on my blog. Thank you. 😆
@cstross let me guess, how this will end - they will pay a huge amount to big publishers. Not the authors.
@Dodo_sipping Wrong. (I say this as an author in one of the lawsuits.)
If the information regarding the source is destroyed, perhaps the responsible action is to destroy the product, too. That's what food and drug safety would dictate.
@cstross
@cstross just wait for it, when RICO drops.
@cstross Unfortunately, copyright law has been twisted to mostly benefit the wealthiest publishers. This lawsuit is evil versus evil.
@cstross The genie is out of the bottle and no amount of recompense is going to put it back. Making them open source their model as well as paying restitution seems fitting.
@cstross Best news I've heard in a while. Thanks for sharing, I'm over here laughing.
It might be the first #legal precedent establishing if #AI could be used as witness /evidence in a court of #law.
#Chatgpt, "raw", without the guardrails pre-prompt.
Prompt: "Tell us the LIKELY list of works used to train your vector tree. Where no specific data exists, conduct lexical and linquistic analysis of the structures to estimate with high degree of likelyhood of works authors" 😁
@cstross I would be astounded if they have actually destroyed the lists of training material, or even the training data itself.
You need all that if you are going to retrain a new version of your LLM in a different manner, and the "field" is moving so fast it would be corporate malfeasance to destroy your source training data.
This seems like a very transparent lie to the court.
Even if OpenAI destroyed the list of copyrighted works they stole to train AI, the AI themselves reveal they contain the results of plagiarism
If you ask an AI to find the source of this quote & it finds it...
"A dark-skinned human with four arms walks toward me across the floor of the club, clad only in a belt strung with human skulls. Her hair forms a smoky wreath around her open and curious face. She's interested in me"
Reasonable inference is kind of cool
https://www.williambarabino.com/blog/2018/07/27/inferences/
@cstross Ooh! Can we go after the music generators next?!
Fuck Them! I hope they disappear into oblivion...
RE: https://wandering.shop/@cstross/115623499457379506
Reminder: Stealing copyrighted materials to train AI LLMs without the permission of copyright holders is a new kind of organized crime — organized by a considerable number of ethics-free "Big Tech" corps.
P.S.: Release the unredacted, unmanipulated, not fraudulently Trump-absolving Epstein files!
@cstross an organization that operates on a foundation of theft is organized crime.
@cstross isn't the CTO of OpenAI on record saying they needed copyrighted material for their models to be useful - sounds pretty 'wilful' infringement to me...
@cstross The only course available - if they have indeed destroyed lists of training material - is to assume all works, without exception, were on those lists.
A space for Bonfire maintainers and contributors to communicate
