As you are aware Microsoft recently changes policies for github.
People on the free trier get access to their AI code generator, and are required to offer their work to the same AI code generator for training.
I consider that totally unacceptable. I’m being strong-armed into working for Microsoft’s profit machine without any compensation and I will loose any author- and copy rights on my work.
But certain things, specifically HACS, require a github account. Github used to be free. It is no longer free.
Is it time to find another way? One that doesn’t require handing over your intellectual property to Microsoft for free? (Yes, that is a bit of a hyperbole…)
Don’t sign up.
ALL the other ‘free’ AI’s out there are the same.
Ever fill in a captcha with pictures of busses and crosswalks and streetlights, did you know you were training Googles auto-cars when you used them.
And don’t click that. There will be strings attached.
Crawl in a hole in the dark and they won['t bother you.
Using copilot is opt-in, but denying Microsoft access to your code and data for training is restricted to paid accounts.
It’s a bit unclear from a legal and philosophical perspective but it means that even though they do not copy your code, they appropriate and use the intellectual property within your code.
Any current Open Source project on Github went into training copilot. Any Open Source author has already been robbed by Microsoft. I can’t really describe what they stole, but it was yours, they took it and there is no way you will get it back. People will ask copilot and the answers contains your IP and nobody wil know.
“You will be assimilated. We will add your technological distinctiveness to our own.”
That is only partially a joke.
But. My point is… Do we want a situation where paid commercial services are needed to access MIT licensed code? “Permission is hereby granted, free of charge, to any person…” except you need to pay Microsoft money or IP to access it.
If you want something similar in a different industry: Music companies are trying to trick musicians into licensing their music to train AI. Which means that a music company doesn’t have to copy the next Taylor Swift album, they can ask their AI to make one just like the real thing. And that is not theory. People are claiming that Spotify is full of cloned music by fake artists, that sounds just like your favourite bands but the revenue of listing to the stream goes to Spotify, not the artist you think you are hearing. I don’t use Spotify, and never have, so I have no idea wether this is true. But, after the incident with a fake Boards Of Canada track a year ago, I think it is a plausible scenario.
Well, my company expressly instructed us to not use copilot, not (yet) to not use github. I trust they made their research, and I suspect what you say is not correct (or not officially, at least).
I agree it makes 100% sense to move to gitlab, though.
Ideally, a self-hosted one, but I’m not really sure if the foundation (which “owns” the code) has any revenue besides Nabu Casa to support this.
And the foundation was explicitly created to create a clear distinction between the commercial part (Nabu Casa) and the FLOSS part.
