I look forward to a future where many natural interactions are improved through computer augmentation. Unfortunately, for the vast majority of today's computing

I look forward to a future where many natural interactions are improved through computer augmentation. Unfortunately, for the vast majority of today's computing interactions, you need some way of conveying more information than what is naturally present. In any well-designed interaction it needs to be clear what you (the user) can do, and what the state of the system you're interacting with is. 'No UI' only works when you're augmenting a system where these two things are already clear. For instance, in the case of the car door system, you already know that you can open the car with your key, and you can tell when the car door unlocks. When I open a new app for the first time, I don't already know everything it can do. I need to see the interface to know what's possible. And I need to see feedback to know that I'm making progress. I've heard that many Nest owners are actually a bit disappointed in its smart features, as there's no way to tell why it's doing the things it's doing (why did it just make it cold in here?) Without a way of communicating its reasoning, people are suspicious of its "father knows best" recommendations. Even Amazon tells you roughly why it's recommending something to you. And Voice UIs don't count as no UI. In fact, they're often a very poor interface, as they convey information much more slowly and invasively than a visual interface, and there's no a priori way to know what voice commands a system accepts.