The AI/ML platform is interesting because I think a lot of it is shifting over to voice-based, rather then web-based (or mobile/tablet-based) interface. A voice-based interface is more invisible than a touch or mouse-based interface.
Come to think of it, adding in the idea of ephemeralization might be a good way of getting a better estimate of timing for a technology. In such a case, voice-interface (and the AIs that power them) would probably be closer to being the next big thing than VR (based upon ephemeralization).
Both Google Now and GMail/Inbox -- as well as Knowledge Graph in Google Search -- incorporate substantial functionality building on semantic web concepts.
Some of them are built on Schema.org markup, which is essentially a modern, multiparty defined vocabulary semantic web markup vocabulary.
I'm not entirely sure how, but the Viv ontologies grows and change as Viv gets used. What would be interesting is to see how the third-party developer interfaces for Viv will look like, and whether incorporating semantic markup in data sources would make it easier to get your offering on the Viv platform. Assuming AI/MLs takes off as the next big thing (which as was asserted, it's not a certainty, but seems likely), I can see the technologies behind semantic web getting resuscitated as people try to get their offerings onto these AI platforms.
IPFS is a completely decentralized store that combines a lot of older ideas in a way that hasn't been tried before. All content is content-addressable (like git), decentralized (like bittorrent), with nodes automatically discovering each other (dht), and a networking strategy that works with NATs. You can use it to publish content, and that content would be pinned or stored by the users who care about the content.
One of the ways you can use IPFS is a kind of decentralized, immutable, unstructured datastore. So if you were writing an app on that, how do you make sure the data is sufficiently structured for your app? You might need to declare schemas, as well as describing a graph of how each piece of data are related to each other. Once you start trying to use IPFS as a kind of decentralized graph database, someone will probably reuse the ideas from semantic web or reinvent it.
Another part of semantic web that had not taken off had to do with www itself. For example, in thinking about blockchain and ipfs, it's possible to create a fully decentralized data store. But to do it well, we'd need to incorporate the semantic web ... And possibly machine learning. I don't think anyone has yet figured out how to do a fully decentralized AI yet.