This is a pittance — about $3,000 per victim, for having their life’s work copied and distributed to, effectively, anyone and everyone on Earth.
And 90% goes to the lawyers.
Cost of doing business.
In reality proper punishment would be forbidding all models trained on copyrighted content from being monetized.
That would actually hurt since the whole point was to build these models and just pay the fine
I would have something to say, but one of the comments in the article said it better than I could.
Definitely ridiculous.
However, nobody is going to prison for making a backup of an album they own.
Like why undermine your own argument in that way? Weird.
Ditto.
So say we all
Yeah and we sit on our collective asses and refuse to make them fix it.
I wouldn’t be surprised if an author wanted to waive there claim to the settlement and look for more, but arguing for damages is a hard hurdle. Depending on the local laws you may need actual examples, or a measurable decrease in sales. Fuck this ask for forgiveness shit
somethingeverything is brokenFTFY
Statutory damages are 150k per violation and that’s how much music industry often got in court per song. This settlement would cover 10K books at this rate, they downloaded over 7M.
This settlement is 0.14% of what statutory damages would have been.
Oh but those rules are only for poors. Silly you.
Everyone is saying this isn’t anywhere near what they should be getting, and that’s true. But this also opens the door to further legal action from others infringed upon, increasing their likelihood of winning.
If it were me I’d ask the judge to order payment immediately, before this bubble pops and they get nothing. It’s all a house of cards built on sand.
They should have to pay ongoing royalties forever. There is no separating the training data now.
The judge ruled that the training was not a violation of copyright
Is the selling of the use of the device trained on it a violation? Because it it should be. Is the use of the device to make money a violation? Because it should be.
No, not according to this ruling.
The fine here was only because they pirated the books they used to train it on
They should be forced to remove all models that have said stolen content or can receive a fine that doubles in value for each violation afterwards. So if the fine exponentially grows larger the longer they choose to break the law. 1 million first fine, 2 million second. 4 the third, 8m the fourth time. etc, etc.
$1.5B/mo licensing fee
If they had to pay out royalties, or pay licensing fees to everyone they stole from, my whole perspectice on LLMs would be different.
Or at least I would be less inclined to make some of the criticisms I have on it.
training AI chatbots on copyrighted books wasn’t illegal
What?!? This is a disastrous ruling!
Then copyright doesn’t exist anymore either. You can’t have your cake and eat it
Not favorable terms to them but…Gotta settle that suit quick so you can cash in with an IPO, before the bubble pops.
Right now they’re settling at their inflated valuation so it’s all Monopoly money.
How is this not favorable to them? They’re making off like bandits, AND setting a bad (for everyone else) precedent
I agree, sorry for my poor writing. I mean to say, the dollar value and decision to settle at this stage on such a fundamental dispute to their business is atypical, and they only agreed to it because of their absolutely nonsense valuation and the IPO plans.
right now they want retailers to hold the bags when the bubble bursts.
chump change.
