You can already do that in a Tesla. Or pretty much any car in that price category. These days the understanding and responsiveness could be described as “pretty

You can already do that in a Tesla. Or pretty much any car in that price category. These days the understanding and responsiveness could be described as “pretty good”. But still nobody uses it because as long as you are able-bodied, voice control is a shitty interface. It requires you to keep context in your head of what the machine is thinking about (instead of whatever conversation you were already having). It requires you to interrupt any conversation you were currently having. It requires the system to give asynchronous confirmation that it fulfilled your request after you have made the request, instead of clicking while you are turning a dial or pushing a button. Anything with fine grain controls requires a back and forth dialogue, such as the volume problem a sibling commenter described. Sure, a better text model might be more graceful at handling a statement like “my windshield looks too foggy” but that’s really a tiny marginal improvement over what we already have and nobody uses.