Hacker News Reader

Comment by voytec | Hacker News Reader

> if a federal judge finds that OpenAI illegally copied The Times' articles to train its AI model, the court could order the company to destroy ChatGPT's dataset, forcing the company to recreate it using only work that it is authorized to use.

I'd like to see it happening but it sounds unrealistic.

View on HN

If I read 1000s of of NYT articles to improve my writing skills, add then write an article of my own, is that a copyright violation?

The problem with the "a lossy mathematical translation of its inputs is exactly like a person learns" arguments, even if courts don't find them ludicrous, is that people absolutely can and are found guilty of trademark violations when they read thousands of pages of the LOTR and then write a fantasy novel full of Tolkein's character names for profit.

A fantasy novel with Tolkien's characters' names is an evident copyright violation regardless of how it was generated. That's not what's happening here.

No, but the point is that OpenAI expects to be held harmless if someone uses its tools to violate trademarks [unlike if a human employee had been paid to read Tolkein and write a story about Gandalf and hobbits] because its not an OpenAI employee consciously doing it, but just a model that translates its given inputs

If GPT is blameless doing some things because it's a deterministic model, not an agent, then the "it would be okay if a person was taught like this" defence doesn't apply in other areas

Camelot! https://youtu.be/m3dZl3yfGpc

A human is not a machine.

Philosophical Mechanism and other philosophical views with Cartesianist roots would beg to differ.

Laws are certainly up for philosophical debate, but at least in the USA, that debate typically has to happen in the legislature rather than the judiciary.

More importantly, though, most judges are not philosophers.

if you had a subscription, no.

The issue is that you know well enough where the dividing line is between a fully original article you write and one that plagiarizes. LLMs on the other hand don’t have that awareness, and can plagiarize by accident.

I think it's the most natural way it would happen. Pit powerful interests (publishers) against powerful interests (Microsoft/ClosedAI). The only surprising thing is it's taking publishers so long to notice this fresh abuse of copyright at grand scale will cost them.

Which is why OpenAIs GTM strategy involves the biggest players in each industry.

That said, the outcome is unlikely - we have trained AI for more than a decade as ‘fair use’ at this point, it’s the application of the technology that is shifting the perspective, nor the act of training.

Every computer vision system in the world is trained on mostly public data for example.

Furthermore, the LLMs purpose is not to generate news so NYT will have to argue about the value of archive data. Many jurisdictions have thresholds of how much of an original work contributes to the derivative before it would be considered not fair use or plagiarism. Given the size of the datasets - good luck.

> we have trained AI for more than a decade as ‘fair use’ at this point

Fair use is about use. Spellchecking ML, search engine ML, etc. all different than ML that produces content.

They would simply move AI training operations to other jurisdictions where it's legal.

They can train whereever they want, but if they want to make the service available in the US, then US law applies.

Reply on news.ycombinator.com