Algorithms present operators with an opportunity to deploy unique services and stand out from competitors, according to Peter Vetter, president of Bell Labs Core Research at Nokia.

“Networks are mostly standardized. All the networks and IT infrastructure use the same silicon technology, use the same processors. So the way to create differentiation is through smart algorithms,” he said during a recent press briefing. 

Operators currently use algorithms to enhance service quality, increase automation to reduce opex, grapple with increasing network complexity and the diversity of network elements, and make better use of resources, Vetter said.

“Algorithms have been around for decades. They’ve been able to solve very complex problems, increasingly complex problems. The very moment you have a set of instructions that are machine executable, it’s a powerful tool and thanks to the improvement of silicon performance, you can deal with the increasing complexity of problems,” he explained.

Artificial intelligence (AI) and machine learning are relatively recent additions to the world of algorithms and they offer new promises for operators. However, these technologies only work when operators have enough data to train and apply self-learning models against, according to Vetter.

Many IT professionals and carrier engineers are surprised to learn that algorithms can be applied at all levels of network infrastructure, not just the edge or core platform, explained Brian Hendricks, VP of policy and government relations at Nokia Americas. 

AI can detect and predict service degrations, allowing operators to anticipate and respond to problems before service disruptions occur, he said. There’s also “great promise” in the security domain, wherein operators can use AI to detect and respond to threats with automated responses, Hendricks added. 

“The anticipation is that you will continue to see reliance and embedding of the technology, and the practice [will be] deeper and deeper as we move to 5G evolved into 6G,” he said.

Vetter and Hendricks also pointed to AI’s growing ability to deal with complexity in radio performance, including beamforming and response to environmental and other network conditions. “We have a model for radio but it’s imperfect, there are corner use cases that you still want to learn,” Vetter said.

Operators and vendors typically build up skills in AI before concepts are developed and deployed in a trial phase, but attracting talent remains a challenge, the executives said. 

“If networks aren’t cool, then the young talent doesn’t come to networks. And having an understanding of just how important artificial intelligence and machine learning are in this space, and the level of innovation that is happening will be important to attracting talent that helps drive the innovation,” Hendricks said. 

Once data and skills are in place, operators can realize benefits rather quickly, according to Vetter. “One important goal is really the enablement of automation, so dealing with the complexity of the network. It should actually alleviate some of the other skills that are required and enable an easier, zero-touch deployment of new services,” he explained.

Operators can and should walk before they run with AI by applying scheduling algorithms to radio or compute resources, and develop scenarios where it can be applied on a slice-by-slice basis, Vetter explained. 

“You can progressively think of adding in AI, machine learning, and replace the classic algorithms to address complexity and complex interaction between problems,” he said, adding that the next level can allow operators to schedule network slices and compute resources against each other. 

“It’s a proportional approach. It's definitely not all or nothing,” Vetter said.