this post was submitted on 04 Dec 2023
889 points (98.0% liked)
Technology
58303 readers
30 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don't think that's the case. A photographer can post pictures on their website for free, but that doesn't make it legal for anyone else to slap the pictures on t-shirts and sell them.
Because that becomes distribution.
Which is the crux of this issue: using the data for training was probably legal use under copyright, but if the AI begins to share training data that is distribution, and that is definitely illegal.
It wasn't. It is commercial use to train and sell a programm with it and that is regulated differently than private use. The data is still 1 to 1 part of the product. In fact this instance of chatGPT being able to output training data means the data is still there unchanged.
If training AI with text is made legally independent of the license of said text then by the same logic programming code and text can no longer be protected by it at all.
First of all no: Training a model and selling the model is demonstrably equivalent to re-distributing the raw data.
Secondly: What about all the copyleft work in there? That work is specifically licensed such that nobody can use the work to create a non-free derivative, which is exactly what openAI has done.
Copyleft is the only valid argument here. Everything else falls under fair use as it is a derivative work.
If I scrape a bunch of data, put it in a database, and then make that database queryable only using obscure, arcane prompts: Is that a derivative work permitted under fair use?
Because if you can get chatgpt to spit out raw training data with the right prompt, it can essentially be used as a database of copyrighted stuff that is very difficult to query.
No because that would be distribution, as I've already stated.
If it doesn't spit out raw data and instead changes it somehow, it's a derivative work.
I can spell out the distinction for you twice more if you still don't get it.
Exactly! Then you agree that because chatgpt can be coerced into spitting out raw, unmodified data, distributing it is a violation of copyright. Glad we're on the same page.
You should look up the term "rhetorical question" by the way.
So you understand the distinction between distribution and derivative work? Great!