Natural language is even worse than video for quick content consumption: you can’t quickly skip through content that way the can fast scroll through a blog post.
If I am cooking and can hold a whole conversation with my virtual chef,that beats a video though. Talking through ideas at my desk would be great. Talking for navigation while I drive, yes please. Shopping with my headphones on and the device seeing everything I see and making suggestions about deals, recipe options, what's low in your pantry, and so on. I'll take that. I see technology becoming more and more transparent in our life. User interfaces will feel awkward when all you need to do is say "show me a video of cats" and it serves it to your companion screen without ever needing to touch tap any UI. No need to even have a web UI for youtube. Just an API and your device does the rest. It can show you a list of related videos without YouTube themselves providing anything more than the data model.
The cost cutting of not needing sophisticated front ends will be a big driving factor if this is as effective as I think it could be.
Anyways, my head is full of ideas like this.
The issue is... I like(d) writing CSS.