Debian dismisses AI-contributions policy
Debian dismisses AI-contributions policy
Posted May 11, 2024 15:13 UTC (Sat) by Wol (subscriber, #4433)In reply to: Debian dismisses AI-contributions policy by josh
Parent article: Debian dismisses AI-contributions policy
And that is a *complete* misunderstanding. I'm in Europe, and if an AI chucked my code *out*, I don't see any problem suing (apart from the lack of money, of course ...)
Feeding stuff INTO an LLM is perfectly legal - YOU DON'T NEED TO GET PERMISSION. So as a copyright holder I can't complain, unless they ignored my opt-out.
But the law says nothing about my works losing copyright status, or copyright not applying, or anything of that sort. My work *inside* the LLM is still my work, still my copyright. And if the LLM regurgitates it substantially intact, it's still my work, still my copyright, and anybody *using* it does NOT have my permission, nor the law's, to do that.
Cheers,
Wol
Posted May 11, 2024 15:23 UTC (Sat)
by mb (subscriber, #50428)
[Link] (1 responses)
Most of the time the outputs are different enough from the training data so that it's hard to say which part of the input data influenced the output to what degree. But at the same time it's clear that only the training data (plus the prompt) can influence the output.
You can do the same thing manually.
LLMs - apparently with some unicorn sparkle magic - remove copyright, by doing the same thing. How does that make sense?
Posted May 13, 2024 7:07 UTC (Mon)
by gfernandes (subscriber, #119910)
[Link]
Clearly not true.
Debian dismisses AI-contributions policy
The output should be considered a derived work of all training data.
Take Open Source code and manually obfuscate it until nobody can prove its origin. It does not loose its Copyright status by doing that, though. And it's a lot of work. By far not click-of-a-button. So it's not done very often.
Debian dismisses AI-contributions policy