Pointing will still be the way to express nouns as we command our machines; speech is surely the right way to express the verbs.
I found it somewhere in The Mythical Man-Month. While the context1 sets the line philosophically against the idea of a WIMPish interface and pushed for more natural modes, I like the blending of pointing with speech which doesn't commit to total conversational mode of interaction.
Most of the works in acoustic interfaces these days go completely towards one extreme by pitching for full conversational interfaces. These are definitely nice but, at least for whatever I do with computers on a regular basis, they are inefficient, unless I have assured, exhaustive and complete high level abstractions. There is nothing new in the fact that, for a known domain, you can create highly expressive and efficient interfaces which look overall unintuitive.
Interfaces differ based on the amount of attention needed to get the real task done. Asking for weather shouldn't really need attention to be put towards any specific device. Nor should accumulating numbers and counts. These are what we can call hands-in-the-batter use cases where the focal point is a working mechanism, like a vessel full of batter, which is separate from the assisting mechanism, like a laptop listing recipes. The assistance is not inherently bound to a certain mode of operation and thus can be made less interfering. This is the place where we ended up adding a lot of consumer devices in the past and are now making them go away by adopting more natural systems like conversational assistants.
But there are, and will be, use cases which don't have such separations. Mostly in situations involving professionals at work. Unintuitive modes are accepted here as they can be made much more expressive for these domain. As an example consider operating computers completely with a keyboard. This needs knowledge of text-ish capabilities and a little training time which is not something everyone will want to spend time on. While many of these use cases are not really needed if you consider the future, there is a potential at present to sneak in unintuitive acoustic input and feedback mechanisms.
The kind of diversity we see in visual interface design is staggering. Many are experimental to be fair, but the amount of experimentation is much higher than in audio where most of the interesting ideas I have seen are in video games and art installations. Maybe the pool of possibilities itself is much smaller here, but let's try dipping toes around. In this weekend, I will try to make a list of possible acoustic interfaces, no speech, that make sense for my use cases and see if I can get something useful out of them.
I found it under The fate of WIMP: Obsolescence