this post was submitted on 11 Jan 2024
92 points (98.9% liked)

technology

23218 readers
2 users here now

On the road to fully automated luxury gay space communism.

Spreading Linux propaganda since 2020

Rules:

founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[–] Infamousblt@hexbear.net 25 points 2 years ago (1 children)

So it's fine if I use OpenAIs content for free without attribution right? That's the same thing? Glad they gave us permission

[–] JohnBrownNote@hexbear.net 22 points 2 years ago (1 children)

~~only if you run it through your own LLM~~

"ai" (and none of this shit is AI, they should have to change their name) "works" aren't copyrightable so go nuts

[–] Awoo@hexbear.net 2 points 2 years ago* (last edited 2 years ago) (1 children)

If ai will regurgitate its training data then you can perform copyright-laundering via this one neat loophole.

We can move literally the entire internet (which is basically all in their training data) into the public domain.

[–] JohnBrownNote@hexbear.net 3 points 2 years ago (1 children)

sicko-wistful

unfortunately i think these things don't keep the training set, just the set of associations and relations it made by analyzing it

[–] Awoo@hexbear.net 2 points 2 years ago (1 children)

Not true, they will completely and totally replicate their training data. The companies try to prevent this so the method to get it to happen regularly changes, but they do it.

Chatgpt: https://not-just-memorization.github.io/extracting-training-data-from-chatgpt.html

Image AIs: https://techcrunch.com/2022/12/13/image-generating-ai-can-copy-and-paste-from-training-data-raising-ip-concerns/?guccounter=1

I'm not saying this would work and you won't get in trouble for doing it. But it would fuck the system just a little bit.

[–] JohnBrownNote@hexbear.net 2 points 2 years ago

oh wow that 's great lol