I find it strange people are saying facebook's leak was the 'Stable Diffusion' moment for LLMs. The license is awful and basically means it can't be used in anything involving money legally.
Facebook has a terrible reputation, and if they can open source their model, it would transform their reputation at least among techies.
I think the spirit of the Stable Diffusion moment comment is that there is a ton of work blossoming around LLMs, largely because there's a good base model that is now available.
And that's undeniable, IMO -- llama.cpp, vicuna are some really prominent examples. People are running language models on Raspberry Pis and smartphones. Anyone who wants to tinker can.
Now, all the stuff that's built on top of LLaMa is currently encumbered, yes.
But all of that work can likely be transferred to an unencumbered base model pretty easily. The existence of the ecosystem around LLaMa makes it much more likely that someone will create an unencumbered base model. And it seems like that is already happening, for example, the Red Pajama folks are working on this.
The author's timeline makes it clear that they feel it was a catalyst. They're separating out "Stable Diffusion" the software from the "Stable Diffusion" moment.
The community has created their own replacement for LLaMA (Cerebras) with none of the encumberance. Even if LLaMA is deleted tomorrow, the LLaMA Leak will still be a moment when the direction dramatically shifted.
The "people" are not talking about the future of where this software is going. They're talking about a historical event, though it was recent enough that I remember what I ate for lunch that day.
Facebook has a terrible reputation, and if they can open source their model, it would transform their reputation at least among techies.
https://github.com/facebookresearch/llama/pull/184