So let me get this straight:
Github scanned a bunch of open source projects.
They trained an AI with it.
They basically ignored the licenses and tried to shove a "because we are basically copying GPL code, it doesn't mean the result is GPL".
And now they are charging for it?
Ok, serious question now:
MS Github scanned whole repositories to train Copilot. Trying to move your code to another service (say, Gitlab) won't delete what it already learnt from your open source project.
But will MS Github still scan new repositories? Or will it find it "done" and now just learn from active, paying users?
@juliobiason i am sure if it is free to anyone it'll come with shackles you have to put on to use it.
@juliobiason My suspicion is that it will be ongoing scanning.
Still wondering about the GPL though. I'd love a SFLC discovery phase to determine how MS is using the code and if the whole codebase could be considered GPL-tainted by incorporating GPL code. That would be utterly fascinating in a world where I had money to throw at this legal exercise.
@juliobiason Just waiting for the first case to go to court when Copilot replicates someone's code line for line
Hoping for an anti-MS asskicking, fearing a boomer judge making a bullshit decision
@juliobiason why would they only scan repos hosted on their own service? Google or Bing wouldn't get very far with that strategy, would they.
@juliobiason Note that #Gitlab.com is even worse than Github: https://git.sdf.org/humanacollaborator/humanacollabora/src/branch/master/gitlab-dot-com.md
I expect the CIA ran out of LSD anyway.
This is the list you meant? It's fab!
@juliobiason you could write your own code unassisted and then have this look at the code you've written independently and then have this tool provide alternative solutions, or hint at resources that might help you in the future. Kinda like a teacher. A teacher is also capable to do your homework for you, but that wouldn't really help you learning.
@juliobiason This is why Microsoft love open source. They can grift on it.
The value which their getting from selling access to Autopilot is value which they didn't create, coming mostly from unpaid volunteer developers.
The loophole is that copyright only covers exact copies. In general, it doesn't cover remixes.
@bob Which generally, I'd consider a valuable loophole!
But still... Whole issue with punching down vs up...
@juliobiason “Because we are copying proprietary code, it doesn’t mean the result is proprietary.”
Try doing that and see what happens..
@juliobiason copyleft is okay unless it's like AGPL since copilot is a service.
Neither is being commercial (FSF supports selling free software), unless they copied noncommercial code too.
Problems are possibility of GH using mentioned exceptions and I don't think it mentioned used projects.
@sadmin @juliobiason What’s mindblowing is not what MS is doing. It’s how many the #FOSS projects still today continue to willfully empower #Microsoft by hosting their projects on #Github, and forcing all users into MS’s walled garden just to file a bug report. The people theoretically on the side of software freedom ignore the 10s/100s of public forges¹.
@sadmin @forgefed @juliobiason @firstname.lastname@example.org I’m glad you answered his question because I didn’t even see the question (I’m probably on an instance that blocks angrytoday.com). Your msg appeard like a new post with no prior thread. I actually had to visit your post on social.tchncs.de in order to see the thread pieced back together.
@nantucketebooks @juliobiason @sadmin The far right column of the forge comparison list (https://git.sdf.org/humanacollaborator/humanacollabora/src/branch/master/forge_comparison.md) indicates FOSS forges with an “n”, so I would try those. I’m not sure if they pass the LibreJS test but if not then it would mean they’ve not properly tagged their JS as free (not that it’s non-free).
@juliobiason Pretty much. And not for the first time, either: https://github.com/MicrosoftDocs/intellicode/issues/201
@juliobiason @jannik Felix Reda wrote a few words about it: https://felixreda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/
@juliobiason Anybody seen a way to opt-out one's open source projects from being scraped for Copilot?
@juliobiason plenty of things on GitHub have no license and this aren't open source. I'm guessing ms isn't fussed about people suing?
@juliobiason how can we make use of this type of behavior being pushed as acceptable? first thought is something like, doing the same against microsoft with microsoft code, but microsoft code isn't worth using to inspire an AI. next thought is something like, train an AI with other copyrighted material, but that won't really hurt microsoft unless someone can think of a more creative use.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!