When I vaguely heard that "AI is reading books and online articles to study without permission," I thought, well, maybe that's just how it is... but it turns out it wasn't that simple💭
Author Andrea Bartz was apparently shocked to discover that her novels were being used to train AI chatbots🥺✨
Why Did This Become a Problem in the First Place?
AI chatbots become smarter by learning from vast amounts of text, and that included authors' books. In other words, their works were being used as "training material" without the authors' permission👀
Bartz couldn't accept this and eventually filed a lawsuit. This wasn't just about personal anger—it was also a battle that questioned the importance of copyright itself💡
What Was the Outcome of the Lawsuit?
The case was settled with one of the largest copyright settlement payments ever made to AI development companies. Moreover, this ruling could have a significant impact on the future relationship between AI and copyright✨
For example, it might lead to more careful handling of data used by AI, ensuring that the rights of writers and creators are properly protected... that's what people are saying📌
Something I Vaguely Knew About, But Made Me Think
Honestly, I hadn't thought deeply about where AI gets its information from or about copyright issues😳 But hearing this story made me think:
- Having your work used without permission is definitely not okay
- While it's great that AI is becoming more useful, we shouldn't ignore people's efforts and feelings
✨
This might be an issue that not just writers, but we as readers and users need to think about as we figure out how to coexist with AI💭
It feels like such a distant topic, but it made me really feel that we're entering an era where these things are becoming more personal🌸
Hatena
Comments
グレース
So the business cost was $1.5 billion, which is less than 1% of Anthropic's valuation.
サラ
She received a $60,000 check instead of a billion and chose not to continue the lawsuit to demand ethical data collection. Well, maybe that's for the best.
ハンナ
I'm just an amateur, but I feel like the settlement doesn't mean much for authors' rights. They got money, but since no legal precedent was set in court, the AI side can just continue as before.
ハンナ
It just looks like they paid money to end the lawsuit, so personally I find this a very disappointing outcome.
クリス
I think the real solution is to create laws that limit the ingestion of rental works to human reading speeds.
ベン
Otherwise, we might end up with a future where they connect part of your brain to a computer and treat you like a cyborg to make it legal.
リリー
But that creates another problem where only major companies that have been collecting data for over 20 years would dominate, making new entrants impossible.
リリー
And they might have already ingested 1000 years worth of works, but would that be recognized or would they have to start over?
ノーラン
This is like money laundering for copyright. Since they can't directly steal the works, they train LLMs on them and then treat whatever results as legal.
ノーラン
Similarly, those who suck up open source code to create apps also get away without following the original copyright.
サム
LLMs are learning using people's efforts illegally.
サム
All AIs trained on stolen data should simply be deleted.
ロバート
(Deleted comment)
ノーラン
If it were a settlement, I'd like to see the top person go to prison and watch them all shift blame onto each other.
キンバリー
Future writers should use cloud book services instead of physical books to prevent unauthorized AI learning, with licenses that prohibit it.
グレース
But to sue users who scrape, you'd need to prove with browsing data that it's negatively impacting your business.
リリー
Unfortunately, digitized physical books don't have that option.
ベン
Is she the one who used all dashes (—)?
エマ
I deliberately wrote weird research to test it, but OpenAI's models definitely learned it and give completely nonsensical answers about the millennium problem.
ハンナ
It feels like they just paid hush money, and AI companies will continue doing the same things unchanged.
クリス
The ruling wasn't a decisive victory—the judge acknowledged that Anthropic downloaded massive amounts of pirated books and infringed copyright, but also judged that using non-stolen materials for AI training is fair use. The authors' side is fiercely opposing this though.
ロバート
Well, until legislation catches up, there's not much we can do. It's similar to teachers assigning homework using books they bought.
