Telecom Networks Are Changing to Meet Demands of an AI World
Telecom networks are seeing fundamental changes as they're upgraded to 5G and eventually 6G, telecom executives said during an RCR Wireless 6G forum Tuesday. Experts stressed that more than ever, networks must be agile and able to change quickly to address evolving customer demands.
Sign up for a free preview to unlock the rest of this article
Communications Daily is required reading for senior executives at top telecom corporations, law firms, lobbying organizations, associations and government agencies (including the FCC). Join them today!
Michal Sewera, head of cloud-native 5G core operations at Deutsche Telekom, said the telecom industry is in the early stages of a major change, and carriers need to develop a new mindset. Operators must do more than just build a new ecosystem, he said. “Suppliers should no longer be called suppliers, but rather partners, because we are building an ecosystem with partners."
As industry launches 5G stand-alone networks, the emphasis isn’t on the “killer app,” Sewera said. The “key paradigm shift” is that “the entire ecosystem is designed and deployed differently.” Networks are being deployed faster “with more IT-like interfaces and IT-like provisioning, and also native capabilities” that can be used by customers, he said.
Being AI native involves more than just adding AI on top of an existing network, said Alex Jinsung Choi, chair of the AI-Radio Access Network Alliance. “It’s a complete shift in how we design the system,” he said. “The network is built from Day 1 with intelligence at the core, not as an optional add-on.” The goal is a system that’s flexible enough that models can be plugged into the network or removed without delays.
Yue Wang, chief technologist at China Telecom, similarly said the network has to be designed from the start to fully accommodate AI and machine learning. Being AI native is not only about using AI and machine learning to manage operations or integrating new AI models into the network, she said. “It’s about how to design the network in order to better provide the emerging AI services on top of the network.”
“It’s clear that the potential of AI is enormous," said Nirlay Kundu, head of technology standards at Imdea Networks. "But we also know that the costs of building and training AI models aren’t trivial.” There’s tension between deploying “massive, energy-hungry AI models everywhere” and industry goals of cutting costs and energy use, he said. The telecom industry is “a very, very low-margin business.”
“Bigger isn’t necessarily better,” Wang argued. Carriers need to focus on specialized AI models that “are the right size” and “fine-tuned for specific network tasks."
Choi said AI native is more than a vendor-driven “hype cycle” to sell carriers hardware and software. “We live in the hype of what AI can do.” Operators want to use AI to improve energy efficiency and the performance of their networks, he said. They also want to expose AI-as-a-service as a way to monetize their networks, he added. AI makes major energy demands, but there are “different flavors,” he said. “We have to make some sensible design choices” to deploy AI in an energy-efficient way.
The telecom industry has historically sold “large-cycle products,” including broadband, mobility and TV, said Ali Tizghadam, a technology fellow at Telus. “Those services are provisioned once and around for years,” he said. “Manual or semi-manual operations were enough for us." Services are now “dynamic and contextual,” he said, citing the creation of a network slice for a two-hour sporting event or a gaming tournament. “They are not set-and-forget types of services.” Rather, they’re “ephemeral” and “high-value,” and having AI-native infrastructure is a necessity.
Ken Figueredo, an adviser More-With-Mobile, said telecom carriers share similar standard components. “For the 6G era, there needs to be more emphasis on software platform components and concepts.” The industry can learn from “market-making software platforms” from companies like Airbnb, LinkedIn and Uber. “They cater for mass markets, but they also cater for long-tail demand,” he said. “They do it on a massive scale,” similar to what carriers are trying to do.
Carriers have yet to figure out how to monetize 5G, and now they're talking about 6G, said Philippe Ensarguet, vice president of software engineering at Orange. “Are we really sure that we all know, at scale, how we could really leverage and monetize 5G?” he asked. “We all know that if we want to address the developer market, we need to talk and live like the developer.”