Crossposted from https://fedia.io/m/[email protected]/t/4218790
TL;DR: In the last year, the Wikimedia Foundation has fired several union organizers, including those that worked on the Community Tech team - a team dedicated to building features for the volunteer community that edits Wikipedia.
As the Wiki Workers Union tries to get the Wikimedia Foundation to recognize their union, it is worth remembering that this is not the first time that the Foundation has worked against the community.
Wikimedia Enterprise is a betrayal of the volunteer movement community of Wikipedia editors, as the Wikimedia Foundation is providing privileged access to big tech AI companies to the Wikipedia corpus - a body of work that the Foundation does not own.
Movement volunteer communities contributed to Wikipedia under copyleft licenses - licenses that work to ensure that the work remains free (as in speech). The big tech AI companies do not license derivative works under copyleft licenses and often do not even attribute where the works came from.
This means that volunteers are working for big tech for free, and the Wikimedia Foundation is selling privileged access to that free labor.
It is against that backdrop that the current unionization struggle unfolds.
The minute Wikipedia allowed AI to train on its knowledge base I stopped donating to them… what a sad sad world we are living in
As opposed to not allowing them to do it and having every AI company scrape the site 24/7 anyway?
They are sitting on $296M. Why aren’t they suing the AI companies to defend the contributors?
Eh, as an editor I haven’t noticed I’m being attacked by AI companies. WMF would need some better argument if they’d want to sue successfully, especially aginst companies that are sitting on 100x more money than them.
BTW it’s already established that training AI models on copyrighted materials is legal without approval of the copyright holder, and IMO it’s unikely WP’s material would be an exception.
BTW it’s already established that training AI models on copyrighted materials is legal without approval of the copyright holder, and IMO it’s unikely WP’s material would be an exception.
That isn’t accurate - this is a highly unsettled question, and there are multiple cases in litigation today.
WMF would need some better argument if they’d want to sue successfully, especially aginst companies that are sitting on 100x more money than them.
A better argument than what - that they are openly violating the licenses under which the encyclopedia is distributed? What more do they need?
I think people have generally (incorrectly) seen certain settlements and deals, like with Disney as “this is legal now” when really it hasn’t been established yet. It’s just gone away for some companies because money.
I genuinely think Disney licensed to Sora because they saw how unprofitable it was and knew it would go under anyways.
Because Disney got OpenAI to pay for a license they now have no use for.
A better argument than what - that they are openly violating the licenses under which the encyclopedia is distributed?
That’s a bit more convincing.
this is a highly unsettled question, and there are multiple cases in litigation today
I followed Kadrey v. Meta a little bit, and the conclusion was in favour of the training being fair use. What are the other ongoing cases?
When there’s a dispute, courts consider the following four issues in deciding whether a use is fair use:
- why the party used the copyrighted material (for instance, for commercial versus educational purposes)
- whether the copyrighted work is informational or for entertainment
- how much of the copyrighted work the party used, and
- whether and how the use affects the market for or value of the copyrighted work.
- Commercial
- Informational
- All of it
- Definitely does
Based on this, not sure fair use holds up that well, but I’m not a lawyer.
Regarding the first point, AI companies usually claim their work is scientific/educational. #3 can be responded to by claiming they actually use none of the original work, i.e. they don’t reproduce any of it; they liken training AI to learning from the materials, and usimg knowledge from some book isn’t just ‘fair use’, it’s the expected use of the book.
It’s mostly bullshit and sophistry, of course, but I’m afraid they will win in most of these cases since they have more money to dump into top-tier lawyers.
I mean, yes. With this type of argument you could rationalize away criminalizing or forbidding anything.
Sure, train models on copyleft data as long as all the training data and pipeline code to train the model is open sourced under the terms of the license.
<!–StartFragment–>
Seemingly, like other community sites, they think that people will continue to contribute to the slop machines for free, probably because they are losers.
<!–EndFragment–>
that is absurd. i know someone personally who just became a wiki editor volunteer. just like any corporation, they treat their lower level staff and volunteers like shit.
