Articles like this promotes extreme views while we should be moderate on what should be OpenSource or not.

First of all, nobody must force OpenSource. Period. Second, such exterme views spawn extreme views in the opposite side.

Discussion is good, but this article is too draconian to be practical.

I don't see extreme views in this article. When LLMs take creative content from copyleft code, the license says the output has to be copyleft. Otherwise, it's theft.

If LLMs weren't trained on AGPL/GPL/LGPL code, it'd be different. But they are.

Taking all of available human knowledge, knowledge that was published under limited licences, and commercializing all that already is a fairly extreme action. Merely arguing that perhaps the fruits of that harvest should also be available to the public doesn't seem that extreme.

IMO it'd be better if the weights themselves were considered public domain.

That’s not an extreme view. Anyone who uses software under a copyleft license should also release any derivatives under the same license. While this isn’t currently the case legally since the GPL was never written with the possibility in mind that LLMs might be trained on it in the future it still aligns with the spirit of the license.

> such exterme views spawn extreme views in the opposite side

The opposite side had exrtreme views for at least 4 decades. They publish stuff, cram it into your face with advertisement and demand purchasing a license for consuming it. And if you don't buy they want you to pay thousands of dollars or go to jail. That's as extreme as it goes. Eff them. Reverse engineer, clone and open source everything.

  nobody must force OpenSource

The current trend is for model output to be public domain, which is “more extreme”. Merely copyleft is not as extreme.

>First of all, nobody must force OpenSource.

No, GPL and copyleft licenses very much do and should force OpenSource by its nature. If someone uses GPL or AGPL and doesn't open source, they should be forced to Open Source.