 Hi, this is Peter Burris with another Wikibon Action Item Quick Take. George Gilbert, everybody's talking about AI, ML, DL, as though, like always, the stack's going to be highly disaggregated. What's really happening? Well, the key part for going really mainstream is we're going to have to have these embedded in the fabric of applications, high volume applications. And right now, because it's so bespoke, it can only really be justified in strategic applications, but the key scarce resource of the data scientists, we can look at them as the new class of developer, but they're very different from the old class of developer. I mean, they need entirely different schooling and training and tools. So the closest we can get to the sort of completely bespoke apps are at the big tech companies, for the most part, or tech-centric companies like AtTech and FinTech. But beyond that, when you're trying to get these out to more mainstream but still sophisticated customers, we have platforms like C3 or IBM's Watson IoT, where they're templates, but you work intensively between the vendor and the customer. It's going to be a while before we see them as widespread components of legacy enterprise apps because those apps actually are keeping the vendors and the customers busy trying to move them to the cloud. They were heavily customized. And you can't really embed the machine learning apps in those while they're sort of so customized because they need a certain amount of data and they evolve a certain type of data and they evolve very quickly, whereas the systems of record are very, very rigid. All right. Once again, thank you very much, George. This has been Peter Burris with another Wikibon Action Item Quick Take.