My current theory: they need more programmers making stuff they can sell for iOS. Popularizing Swift is beneficial to them because it still technically keeps people locked in the Apple ecosystem (a kind of soft lock-in - they've opened the language, but it's still primarily supported by Apple).

Increasing the pool of hirable developers does not otherwise exclude selling computing toys to the general population.

I haven't heard of Swift Playgrounds for iPad before. I've just finished watching its demo - I must say, it looks nice and has some cute solutions for entering code via touch interfaces (like that for-loop dragging thing) that I hope will spread around to other applications. That said, it's an educational app - I can learn some Swift with it. But what can I actually program with it? Can I use it to make my calendar talk with my SMS app? Or with my Bluetooth headset?

That's the thing I complain about when I say that mobile devices are developed as toys, not tools. You're always limited to some functionality provided by the vendor and third party apps. Apps that don't talk to each other, that restrict your choices to few operations their authors thought about. Not to mention apps that increasingly want to suck out and monetize all your data, but that's beside the point. On Android we have Tasker, which is something that I believe should be a default component of Android (albeit it could use some ideas from Swift Playground to make UX better). But it isn't, and the current trends in mobile, web and desktop suggests it'll never be - the end user is forbidden from using the computer, they must only ask their apps for services.