> just be a continuation of a trend, where the effect is already in full effect with slightly older kids, who can read

There's more to it.

I'm interested in multi-scale interdisciplinary story telling for science education. Sort of, sit down on the floor with a graybeard, who has fascinating stories about engineering and the world. And videos and interactives and questions. And you "pick up" all sorts of understanding and skills. Rather than the usual incoherent memorize and regurgitate.

So say you're today writing a web page with embedded videos. When showing the page in person, you can see what's of interest to the viewer; when something is getting boring; what is understood and what's confusing. You can respond to questions, address misconceptions, point to related content, and suggest different paths through the material. But what about when the page is viewed normally? Youtube-like "related links" are very limited. You can, at long last, do speech recognition, on some browsers. But dialog systems? Not ripe. Something as simple as "if they ask about those ripples, show them this other video" still isn't easy enough. "[R]espond in a meaningful way"? That could be breathtaking.

Someone did an art project. You go, and watch an interesting video. But why is it interesting? Under the covers, it's like an old "choose your own adventure" game, assembling a narrative out of parts. And it uses eye tracking to detect what you're interested in. Do you like character A? Then we'll emphasize A.

There's old advice, that since history is all connected, you should approach it from whatever interests you. Textiles, architecture, materials, food, games, environment, governance, whatever. It's all tied together, so start pulling the thread anywhere. Science and engineering are also very very richly interwoven. Or should be. We just teach it so very wretchedly.

And then there's AR/VR coming.

And for the "but irreplaceable humans!" arguments, there's work on collaborative human/crowd/computer composite agents. Like http://hdl.handle.net/1802/29594 .

So yeah. The next decade could be awesome.