阻礙應用程式在使用者裝置上本地運行大型語言模型伺服器的因素有哪些？

I was using Gemma’s models in the plane offline while reading a book, to aid my studying, and I thought about this. Think about it this way. Most phones can run Gemma’s 2B model (many the 4B too), and open source models will get cheaper and better, and CPUs more optimized for AI. Gemma 4B is almost on par with GPT 4o which was the #1 model at its time. I think in the future, it will work this way: Client: Request -> Local LLM using phone as server -> Server: Response -> Phone view (client)