With a model licensed using MIT, you know what you are getting. You know you can modify it. You know you can use it for any purpose whatsoever, including ones the corporation that built it dislikes. Using the MIT license is "no ace up my sleeve" licensing - there are no gotchas that will come and bite you later.

The EU restrictions were added in 3.2's Use Policy, while 3.1 License introduced the restrictions on using model output for training other models. I've collated the changes here: https://notes.victor.earth/how-llamas-licenses-have-evolved-...

Besides that, I agree with what you're saying.

No indeed, full stop. I think that If one needs to express some idea, one should use clear and exact terms for this. I don't think "open source", which is widely recognized with a specific meaning, should be twisted. The shared definition of "open source" is the one defined by the OSD.

> But for the majority of people, they don't care

Maybe they should, but even if they don't, it's not an excuse. Especially in an article about the very topic at hand.

The open source definition seems constantly attacked for some reason, and that blurs the meaning of everything in the ecosystem. We need to fight back against this. If you care about open source, how are you supposed to spread the word if "open source" has lost its meaning?

It also matters because if one thinks that open source is the better thing to do, but everything and their dog is called open source, then everything is good and we don't need to bother with the principles anymore (not unlike the Overton window).

Now, in this case for this article, it's just an example that doesn't work, it's not something fundamentally wrong about it. I'm just pushing back on the idea that it doesn't matter if we misuse "open source" a bit which seems to spread fast these days.