I think this is a big issue with voice-controlled UIs generally. People will certainly not try to say anything that comes to mind because they know the "assista

I think this is a big issue with voice-controlled UIs generally. People will certainly not try to say anything that comes to mind because they know the "assistant" is a machine with limited understanding of the world. But how do you allow them to discover what's possible, then? (I've never used one of these, so I don't know how this is being done today, if at all.)