Yes and no. Everybody to some degree needs open models based on public data. That's because part of the power of these models is that there needs to be a gene

Yes and no. Everybody to some degree needs open models based on public data. That's because part of the power of these models is that there needs to be a general-purpose foundation that the LLMs can be fine tuned on for specific purposes. We teach children a variety of subjects because, if you train a child to do nothing except weld for 12 years, you might get an adult who can competently perform very specific welds but is otherwise a bumbling illiterate idiot who can't take custom instructions or do basic arithmetic when just holding a torch isn't enough. You might be able to train an LLM from scratch on a very specific corpus, but LLMs only really become useful when they have an amount of general purpose knowledge. Plus there's the problem that training an LLM from scratch takes a ton of time. But the siloing of specific knowledge can make a lot of sense. Specialist LLMs may be closed or semi-closed but still be based on a foundation of open data.