In the last six months, I’ve heard operators express concern over the diminishing 5G opportunity, and network vendors express concern over the state of operator spending and the fact that 5G is sinking as a driver. The same two groups are working hard to believe that 6G will solve all problems. Of course, that’s what they had hoped for 5G. The new angle, both groups suggest, is that we didn’t have the onrush of AI to boost 5G and that magnificent force will be there to boost 6G (and operators, and vendors) over the top. OK, maybe that sounds a little cynical, but it’s substantially true insofar as what I’m told. Is it true in terms of outcome?
Yesterday I pointed out that AI could have either a broadly constructive or destructive impact, depending on how enterprises applied it. Telcos are dependent on the mass market, so they’re highly impacted by anything that drives GDP up or down. If AI is to help telcos, it has to be broadly beneficial economically, which means it has to improve productivity and empower and enrich more people than it displaces.
5G taught us something important. You can’t create an ROI for something that doesn’t create “R”. It doesn’t help to list what that something could have or should have done, only what it actually does. So what we know about 6G is that it has to drive spending, and if neither enterprises nor consumers will spend more for the same-in-the-end service they’ve had all along, then 6G has to get its magnificent force from some new application. Obviously, AI has to be it, because nothing else of similar market significance has come along. That sort of inevitability explains the commitment to AI but doesn’t guarantee a favorable outcome, so if we want our good result, we need to identify just what 6G might do to leverage AI, and how to realize that specific thing or things.
Don’t say that AI will increase traffic, please! Traffic is what all network services depend on, with or without AI. I submit that content has increased traffic, but it really hasn’t moved the needle much for willingness to pay, and nobody is rushing to tell me how 5G has exploded in revenue generation for operators because of content. Why would AI be different? It’s not just traffic, it’s the utility of the application from which the traffic is generated.
Don’t say “latency” either. OK, 6G is supposed to improve latency by about ten times versus 5G, but to believe that’s important, we’d need to see a bunch of new 5G latency-sensitive applications inhibited by 5G limitations. Where are they? If latency is to be critical in justifying 6G, some new thing has to make it so.
The point is that when applications or services are constrained by something, they tend to cluster at the point of constraint waiting for relief. If you could expect to sell a bunch of two-gig broadband, you’d expect that current customers would be constrained by current limits and cluster at the top end of offerings. They aren’t. Thus, I believe we can say with confidence that there aren’t any characteristics of network services that are constraining those new revenue-generating applications for 5G or 6G, or any-G. That doesn’t mean that there aren’t any such things, only that something else is constraining them before the network becomes an issue.
Almost everyone carries tech with them in the form of smartphones. To varying degrees, we’ve integrated them with our lives, and more so with each passing year. We hear a lot about how people are using AI do help them do various things, so it’s not unreasonable to think that AI might be the application that would drive traffic, right? Only maybe.
Suppose we need to research a topic. We could visualize two options. We might use a search engine, find the information, download it, and digest it. Or we might ask an AI entity to do the job. In the former option, we potentially use a lot of bandwidth, but we also use a lot of our own time. In the second case, our AI entity might do a lot of data digestion, but the internal research traffic is inside the AI data center, and all that flows to us is the result. So, if AI follows the hosted-service-chatbot model, it doesn’t really do much for traffic.
What we need to do to make greater use of AI into a real driver of traffic we need the concept of an AI entity, but we need it to be on the phone, pulling data on our behalf. This means that the common chatbot/copilot model that relies on cloud-hosted AI isn’t likely to create any mass-market opportunity at all. What would? But at the same time, we can’t assume that this to-the-smartphone information is carried to users from paying 6G sources in personal sessions. Sure, telcos would love that, but nobody would pay for it. We need a scalable solution, whose cost doesn’t explode as usage explodes.
For years, I’ve blogged about a concept of real-world augmented reality (a metaverse, if you like), postulating that an AR-equipped person could move simultaneously through the real world and a virtual world created by “information fields” published in wireless form (WiFi, cellular, whatever) by real-world things like other people, stores, vehicles, buildings, cities, etc.), and could then visualize via AR select information from those fields. This could be augmented by on-phone AI, which could both select what to visualize and assist in the visualization. Obviously, the fields could be created or augmented by AI as well, and could be hosted in a variety of ways appropriate to the latency requirements. And, of course, you could do this without AR glasses, using the fields but not visualizing them.
These information fields cannot be individual sessions, as I’ve already noted. Instead, they’re essentially like electronic billboards that cite what something is, does, or is doing. This information is available to any field receiver in the area, the size of which could depend on the nature of the thing asserting a field. You could use this technology to describe landmarks, offer weather, traffic information, prices on special deals, etc. You could even envision individual vehicles asserting their trajectory and status. My original thinking was to have personal agent technologies cloud- or edge-hosted collect this, which would not have required it be actually transmitted, but I think the shifting times suggest we should think of these fields as being multicast to smartphones, smart vehicles, and so forth. In fact, I think that a suitable multicast model to support this sort of thing is the most critical thing 6G could bring to the table, in large part because it would promote distribution of information to the handset, which empowers both handset and network.
6G standards, as proposed, do include multicast, and the concept is even linked in documents to IoT and extended/virtual reality. What’s lacking is the connection between the service capability and applications, which is what I’ve tried to supply with the example of “information fields”. But, for example, how can the information fields be managed so that they don’t create radio-frequency chaos and collision? Can the industry somehow be expecting outside players to develop answers to that, and other questions? If so, isn’t reliance on unidentified third parties what ensured 5G wouldn’t live up to its expectations? After all, 5G also included multicast.
Absent multicast, AI either centralizes in the cloud, disenfranchising telcos who have to build out 6G, or requires edge computing to host applications that telcos have been unwilling to directly support. It means you don’t rely on smartphones, and don’t need incremental 6G capabilities at all. In contrast, a smartphone-centric approach presents several advantages for telcos and 6G.
First, the whole notion of information fields could become a 6G requirement. Think of these fields as short-to-medium-range multicasts, a reasonable network/service feature.
Second, making the handset the focus of interpretation ensures that traffic goes there and not just minimal results of some sort of inquiry.
Third, any 6G adoption will demand handset renewal, which engages the smartphone vendors and also validates the relationship between phone replacement plans and service adoption/renewal
I am not seeing any real indication that 6G is addressing any of this in a realistic way. All the standards want to do is to offer the same feature support that 5G offered, and that wasn’t exploited. That has to change, or we can’t expect AI to help it, and we can’t avoid having 6G fall into the same traps that 5G did, which was supporting features for which there was no demonstrable willingness to pay. Telcos want to talk about the benefits of 6G, but those benefits rely on driving changes to applications that neither telcos nor their vendors seem willing to undertake.
The Nvidia investment in Nokia to integrate AI with 6G is either an opportunity or a giant waste of money and effort, depending on whether the two can actually help develop these applications. That means thinking way outside the “telcos sell 6G to IoT devices because they’ve run out of humans” box, and actually figuring out how to build a new application set. Are “information fields” the way, as I’ve suggested? Maybe, but we need to look at them, at other options with serious intent to drop the usual bandwitdth-and-connection crap and address real application-level issues.
