It's actually the people who believe that Android is Linux-based that are mistaken. If you write for Android, you write for Android and nothing more. You can't just run the application on a regular Linux (i.e. the one with libc and X11), and if Google ever rips Linux kernel out of it and replaces with, say, DOS, most of the applications won't even need to be recompiled, because they were using Android's API/ABI.
No, not quite, not without enormous effort. Look at Cygwin or WSL or Wine: tons of work poured into and still not there yet.
On the other hand, typical Android application only ever touches Android's API and doesn't even leave its JVM. Replacing whole kernel (with obvious single point of work in Java interpreter) would be practically invisible.
When we can reasonably "chat" with our voice assistants then we have actually gotten there. More specifically, when I can tell my voice chatbot I just want to say "alexa" in the morning when the alarm goes off to snooze. Or when they are adaptable and pattern observant enough that I can say "alexa the usual" in the morning and get the higher volume as I walk through the house and news and my favorite podcast as I ask for every morning. When they come close to accomplishing those things, then we will have vocal chatbots. When they can "remember" just as much as the original Eliza chatbot then we will be there. Currently they are more like vocal light switches than vocal chatbots.
The Voice Assistance is a vocal chatbot. People were thinking it was going to be text.
No one missed anything except the actual product everyone was talking about. I feel like this happens all the time. We get caught into one strict interpenetration and miss the HUGE thing staring at our face, or ears in this instance.