Personally, I think any output from these models must be for noncommercial use only, as its impossible to track down exactly whose rights are being violated for any given output, yet guaranteed that’s its a non-zero number every single time.
They want it to worm its way into every corner of our lives as quickly as possible so that when a non-criminal administration and congress come into power, they’ll be hesitant to prosecute something that touches the entire economy.
To be clear, I said I’d like to think all the output should be copyleft. I actually think that, since the training set is a hopeless mishmash of conflicting licenses, it’s just massive copyright infringement and can’t be used for any purpose at all. Part of me is surprised some judge hasn’t started ordering them to just straight-up delete the entire models yet.
But which? Creative Commons have a range of different restrictions some aren’t even GPL compatible, copyleft as per OSI or GPL would still impede on the source available code licensed under CC BY-NC and vice versa.
And that’s forgetting all the blogs and books where no open license is given.
Exactly. And that’s before adding in all the “source available” code that’s publicly available but not actually Free, let alone the leaked proprietary code that they’ve probably shoved into the training corpus willy-nilly, too.
Biggest problem is proving where the code came from. Trying to pick out what software vibe code comes from is loke.trying to pick out a single hay strand from a pile of sharp broken glass and used needles. Can you do it? I’m sure with enough time you can, but at that point was the time spent worth it instead of learning how to be an actual programmer and/or developer?
and this is why AI code should either be considered public domain or illegal by default, unless the user can prove which code went into the training data.
…Hey, wait a minnit! That’s not a bright side at all!
(Personally, I’d like to think that the output of any LLM whose training data includes copyleft code ought to be required to be copyleft.)
Personally, I think any output from these models must be for noncommercial use only, as its impossible to track down exactly whose rights are being violated for any given output, yet guaranteed that’s its a non-zero number every single time.
They want it to worm its way into every corner of our lives as quickly as possible so that when a non-criminal administration and congress come into power, they’ll be hesitant to prosecute something that touches the entire economy.
Don’t let them win.
To be clear, I said I’d like to think all the output should be copyleft. I actually think that, since the training set is a hopeless mishmash of conflicting licenses, it’s just massive copyright infringement and can’t be used for any purpose at all. Part of me is surprised some judge hasn’t started ordering them to just straight-up delete the entire models yet.
But which? Creative Commons have a range of different restrictions some aren’t even GPL compatible, copyleft as per OSI or GPL would still impede on the source available code licensed under CC BY-NC and vice versa.
And that’s forgetting all the blogs and books where no open license is given.
Theft is theft. AI CEO’s should be in jail.
Exactly. And that’s before adding in all the “source available” code that’s publicly available but not actually Free, let alone the leaked proprietary code that they’ve probably shoved into the training corpus willy-nilly, too.
Biggest problem is proving where the code came from. Trying to pick out what software vibe code comes from is loke.trying to pick out a single hay strand from a pile of sharp broken glass and used needles. Can you do it? I’m sure with enough time you can, but at that point was the time spent worth it instead of learning how to be an actual programmer and/or developer?
and this is why AI code should either be considered public domain or illegal by default, unless the user can prove which code went into the training data.
I think you’re right. Hopefully FSF will win some lawsuits and we’ll get all the models made public
Indeed it would/does