@Michael Stanclift How long until GitHub turns off the Mastodon foundation's Container Registry account unless they start paying?

I think they already do.

GitHub has no any limits for public projects. But this "how long until..." can be said to any service. At least GitHub has always been supportive for open-source projects and there is nothing which indicates otherwise at the moment.

@nicce In my book, the Copilot project is a massive fuck you to all open-source projects and the sale to Microsoft was the final straw. I can't recommend GitHub for any open-source project anymore.

The fault is more like in the current legislation. There are no proper laws how A.I can use licences and what are ethics. ChatGPT et. al. have for sure used the same public resources as Copilot.
Microsoft has let the GitHub function on its own, so the sale has not have much impact yet but on the other had has helped to keep the pricing low (and free public use) for sure, since it is backed by such a big corporation.

@nicce I feel like the "yet" in your reply is load-bearing. I'd rather a paid service than a free service that is financed by harvesting data that definitely wasn't submitted for commercial purposes.

And mentioning ChatGPT and other LLM systems will not make Copilot look better in my eyes. There all are awful in the data collection department.

"a paid service than a free service that is financed by harvesting data that definitely wasn't submitted for commercial purposes."

The problem is that there are no many other sustainable business models "yet", which would keep infrastructure up on scale and keep adding new features. In that sense, the "free" indeed is misleading. You are paying with your data. Something which is truly free, is hard to find. Somehow the infrastructure and the time of people need to be funded. Otherwise, people can't keep doing that since they need food and house with roof. There is Codeberg, for example, it could not sustain the amount scale GitHub has, nor keeping adding new stable, features.

@nicce GitHub had paid Entreprise features even before the sale to Microsoft, so I don't believe exploiting this data was their last resort to recoup expenses related to their free tier services. Even then, they should have collected the informed consent of all the projects they were intending on using the code as training data for Copilot. If this sounds unworkable, it's because most current LLMs are simply breaching consent and including material they have no business using in their model. No benefit is worth breaching consent at such a scale.