> That said, I don't agree with the dig at LLMs, ... [it] is odd in the context of open source. I don't see it that way, if talking about the stricter end of O

> That said, I don't agree with the dig at LLMs, ... [it] is odd in the context of open source. I don't see it that way, if talking about the stricter end of OSS licensing. There is an argument for training a model with AGPL code meaning that the resulting model should be released in full to its users as a derivative work, for instance. Being an ideological complaint doesn't make it an invalid one, even if you consider that ideology to be rather dogmatic.