Talk about how you're embedding artificial intelligence and machine learning into your products.
Certainly in the PC domain we do that, and we do it in the printing domain. And increasingly in the 3-D space. I'll give you a couple of examples. In the PC and personal system domain, you essentially are collecting -- with appropriate privacy and permissions -- information about the product itself and the performance of the product. And you're using it to model the overall performance in the product, so you can predict failures before it happens. And then you can do things in advance in order to mitigate the pain of a potential failure. And we do that in the print domain as well, where we can keep track of usage. And we use that then to offer differentiated services, like Instant Ink. So that we know how much you print, and we can use that to determine when you run out of ink. And then you can sign up for a service where you don't ever run out of ink.
Now going forward it will get a lot more complicated. It will get a lot more sophisticated. And you can see it'll start getting deployed in other areas -- 3D print as well, where we'll use machine learning to do things like thermally pack a print bed with the optimal packing in the bed itself.