Years ago, I angered some people by calling a new technology a “UFO”. Why did I do that, and why did it make some people mad? And finally, what does it have to do with network traffic and telco revenues? Read on.
What’s the primary property of a UFO? I contend it’s “unknowability”. As I said in that old talk, one of them isn’t going to land in your backyard, so you’re free to assign it any properties you find attractive without fear that reality will stand in your way. I made the association with a new-at-the-time technology because the online world was full of buzz about all the changes that technology would drive, when in fact it wasn’t doing anything at the time. Why did that make some people mad? Because they wanted to believe in their UFO.
A Light Reading piece captures this dilemma. The story cites multiple sources that believe that AI will generate massive growth in mobile traffic. “I can’t put my finger on it, but what I would say is, as generative AI moves to the mobile device, you’re going to see a massive pickup in mobile data traffic,” says one source. The UFO is still circling and spitting out promises. Another source (Akamai) that says that data consumption growth has hit an all-time low. So the UFO may be landing after all, and its properties aren’t too attractive.
AI doesn’t generate traffic. No technology does. What generates traffic is money, opportunity, missions. The reason why Akamai is seeing the slowest rate of growth in broadband data is that there’s a limit to what we can consume, based on its value to us and its ability to deliver money to the stakeholders needed to do the delivering. Every opportunity saturates eventually, so you need new opportunities.
What new technology can do is to help unlock those opportunities, but you can’t jump from technology to traffic and skip the opportunity step. What does AI unlock? What would the cloud, or IoT, unlock? How does the unlocking work, and what’s the total benefit the opportunity can deliver? Who all are the stakeholders, and how does each get a share of the benefit to justify their participation?
Let’s apply this to the question of mobile traffic and AI. What would generate mobile traffic? Answer, something that uses mobile service. That would mean a portable device like a smartphone or tablet, or a sensor/controller in an IoT application that was based on a cellular radio. So, for AI to drive mobile traffic it either has to increase the traffic from a portable device or drive cellular IoT.
The idea that generative AI would move to a mobile device needs some exploration. Does that mean that generative AI is used from a mobile device, or that it runs on one? If the former, then why would mobile AI be any different from fixed, since both would be accessed via the Internet? We have not seen any pickup in Internet traffic because of generative AI. If we’re talking about running generative AI on a mobile device, isn’t that a bit optimistic given that the same article recounts the billions that companies like Meta and Microsoft are spending on AI hosting? Gee, tell them to use their phones instead? C’mon, people.
How about cellular IoT? That’s a reality in the same sense that satellites are a reality, but most people don’t launch their own. Your home and office are likely to include IoT devices that communicate, but the great majority don’t communicate via cellular service. Those that do are probably security systems that have cellular connectivity to the alarm center. Do you expect AI would create more break-ins? If not, then we’d need to see some IoT missions that could not be satisfied using wired devices or WiFi.
I think you see the tip of problem here. What has to be developed, and promoted across all the necessary stakeholders, is the new missions. The technologies to be used are products of that promotion and depend on its success. Deeper down, the problem is that the next steps that we need to take to move tech forward are harder than they’ve been in the past. It’s the nature of us all to “pick low apples”, meaning to take on projects that are easier to justify and complete. The inevitable result of this is a bunch of trees that are bare within our reach, trees whose fruit requires more work, even a lot of work, to reach. That’s where we are today, and we’re not going to change that by just introducing a new technology. It has to build business cases, and that means we have to build them. Do we believe that AI is going to create its own justifications, figure out how it gets integrated into business operations and our daily lives? But enterprises say that they can’t do this because they can’t acquire and retain skilled AI types, can’t compete with Silicon Valley.
There’ some good news here, though, and it comes from (of all places!) the enterprise response to the Great Cloud Repatriation Story. While few enterprises really tried to stamp out cloud usage, or even to completely remove applications from the cloud, they all have been struggling to do a better job of planning applications around cloud cost/benefit realities. About a tenth of those I chat with have come up with what seems a workable approach. Instead of trying to hire developer experts on the new technology (the cloud, or AI), you get a mixture of software and enterprise architects to examine the business processes that have to be optimized to make a project business case. The goal is simply to build an application model that presumes the cloud or AI features can be exploited, then when the model is built, let developers examine the available tools. That, this group of enterprise say, is within the capabilities of their current development teams.
Network traffic is a product of the value of moving information, and new traffic means new information values. If AI, or the cloud, or IoT, can create them, then it will pull through network traffic gains. If we don’t want that dazzling new technology to be a UFO, then we have to bring it down to earth for exploitation. The ten percent of enterprises who seem to be getting things right are already reporting that their new projects are in fact increasing network traffic, but it’s still early to say how much the increase will be and what impact it will have on LAN and WAN, on mobile and fixed services. But the network has always been the tail, not the dog. We need to wait for information value to drive information movement.
I think this experience shows that we need to recognize that the goal of IT isn’t to consume technology, but to improve operations. To do that, you can’t start with the technology, you have to start with what you want it to do, and then pick from both leading-edge stuff and traditional stuff to implement your plan. The barrier to this may be that it’s not the sort of exciting thing that everyone wants to read about, hear about. One CIO, with a touch of bitterness, said “I can find a dozen stories on artificial general intelligence, but none on how I can actually use AI.” I think those stories will come, from the ten percent of enterprises who seem to be getting it right, but we surely could use some realistic early PR to move things along.
