1) Talent. The more engineers are familiar with Llama, PyTorch, and the likes, the easier it is to find ML talent for Meta.
2) Free research and innovation that Meta can readily copy for their needs.
3) They don’t really give up any power. They can always keep future models proprietary.
In my opinion, it is a bit like Microsoft’s “Embrace, Extend, Extinguish.” Make SW publicly available, extend it to your own business needs, and if a competitor emerges, the power is still with Meta to restrict its use or copy the use case (“extinguish”).
2) Free research and innovation that Meta can readily copy for their needs.
3) They don’t really give up any power. They can always keep future models proprietary.
In my opinion, it is a bit like Microsoft’s “Embrace, Extend, Extinguish.” Make SW publicly available, extend it to your own business needs, and if a competitor emerges, the power is still with Meta to restrict its use or copy the use case (“extinguish”).