They don’t and they don’t need to. It simply is something that proves the quality and capabilities of Meta AI. They use it themselves, see improvements of it like Vicuna and Alpaca and can incorporate those improvements instead of doing all R&D themselves. Furthermore, very large companies will buy other AI models that have commercial versions from Meta AI, because Meta has shown it is an AI powerhouse. So they don’t need to make money in that sense, as Meta also does not have any real additional cost now that LLaMa is trained. But they also still make money, by having it incorporated into Azure and AWS, where you pay per computing unit and hour and they most likely get royalties from it that way.
Latest Answers