Telco AI: Critical for networks, but not yet quite a business
Artificial intelligence is already a critical tool in running telecom networks – but operators are still a little way from being able to monetize it.
That's the message from Thursday's AI session at MWC Shanghai.
China Mobile is one that's taken advantage of AI to operate its enormous infrastructure.
Executive Vice President Li Huidi says the telco giant has developed a tool to optimize its intelligent antennas, each of which has 10,000 potential parameters.
"We are using AI clustering to optimize the algorithm," he said. By crunching coverage attributes and historical traffic volumes, it can automatically configure each antenna.
The result: an 13% increase in outdoor downlink speeds and a 30% boost in indoor speeds, Li said.
AI is also helping China Mobile maintain its network, which generates a staggering 100 million alarms daily.
The network can now recognize 20 different kinds of fault scenarios and can send out 2000 alarm responses, Li said. "Everything is visible, controllable and traceable."
In future China Mobile would make sure AI is embedded "into all aspects of production."
For vendor Ericsson, "AI is a very big toolbox," according to research director Elena Fersman. Network automation had become essential because of the growing complexity.
She said Ericsson applied AI to deal with complex network use cases like huge traffic volumes, massive IoT and the connection of more and varied devices to the network.
She also had some metrics to share. For example, RAN KPI degradations, an important measure of RAN performance, had fallen 70% through use of predictive AI.
Autonomous fault detection had driven down MTTR (or mean-time-to-response) by 90%, while energy-related opex has been cut by 15% and intelligent RAN design has brought about a 10% saving in RAN capacity investment.
In contrast to the advanced use of AI internally, telco efforts to generate revenue from AI-based capabilities are still a work in progress.
China Unicom offered a list of AI-enabled use cases delivered on its recently launched industrial Internet platform.
These included quality assurance checks for factory goods, various kinds of compliance checks for staff, and the use of remote-controlled robots to monitor factory floor safety.
But for now, it's still just a catalog. Chief scientist Lian Shiguo said Unicom is offering a trial service for customers, dubbed "buy after try."
"We really want to do validation of this agility," he said. "We want to empower our partners."
— Robert Clark, contributing editor, special to Light Reading