I see. However, about speech recognition for instance, it is for now done "in the cloud": your recorded voice is sent to a server which does the intensive comp

I see. However, about speech recognition for instance, it is for now done "in the cloud": your recorded voice is sent to a server which does the intensive computations and sends the corresponding text back to your phone. Why would that change? It even allows Google or Apple to train their speech recognition system so I think they will want to keep it that way. More generally, more and more computation are done in the cloud, even on the desktop (Chrome OS). That said, I agree about high-res gaming and augmented reality, etc.