-
AWS surveyed 100 telcos on what they thought were the main AI use cases
-
Respondents said network use cases for AI were the highest value
-
One network use case is to optimize massiveMIMO radio beams
CONNECT(X) ATLANTA — Connect(X) used to be mainly a cell tower show. But things have radically changed. This year, one of the first keynote panels in the main ballroom focused on generative artificial intelligence (GenAI) within telecom.
“I work with telcos across the world. When GenAI burst on the scene early last year, we went out there and surveyed over 100 telcos on what they perceived were the main use cases and low-hanging fruit," Ross McWalter, head of Telecom Applications at Amazon Web Services (AWS), said.
The most obvious initial use cases specified by respondents were natural language chat for internal purposes. “They wanted to learn before they moved to external use cases,” said McWalter.
And of course, GenAI chat for contact centers was of interest to the telcos that were surveyed. “It’s generally a huge cost center, and you can get very quick operational cost savings there,” said McWalter.
Telcos that participated in the AWS survey also expressed interest in using GenAI to help sell their products.
McWalter said, “If you look at Netflix, 80% of usage is driven by personal recommendations. If you can get it right, you can certainly drive up revenues.”*
Network use cases for AI were deemed by survey respondents to be a lot more complex but of the highest value, he said. AWS is currently looking at network mean-time-to-repair use cases in 2023.
Network uses for AI
Gerardo Giaretta, VP of Product Management with Qualcomm Technologies, said the chip company is seeing a lot of initial uses of AI for network planning and digital twins. He said the digital twins technology is “really continuous automation.” It constantly processes data from the radios and core network to provide an “almost real-time view.”
Qualcomm's Edgewise solution basically does orchestration of radios, Giaretta said. It can provide predictions of traffic loads to prevent surges or faults. Edgewise feeds on the data that comes from the radios and the core network.
AI can also assist with how massive MIMO radio beams operate, Giaretta said.
On the sidelines of the show, Fierce Network spoke with Mobile Experts analyst Joe Madden, who said that AI can optimize massive MIMO beam steering by organizing radio loads and user groups.
For instance, AI can identify a group of users who all happen to be traveling on a highway in the same direction. Then AI can optimize the radio beam for those users.
Madden said, “If you took at a typical 5G network it can deliver data for $.25 per gigabyte. If they implement AI and get a 25% boost in capacity, that can reduce the costs per gigabyte to $.21."
Past and future
Wireless and wired telecom operators have been virtualizing and cloudifying their networks for several years now. Qualcomm’s Giaretta said, “I think that movement is a first step to bringing AI into the network. Many operators are going down this path.”
“Over the next 12 months, I think we’re going to see a lot of these uses cases move into production. And there will be a proliferation of AI and GenAI across all areas of the business and value chain," AWS’ McWalter said.
*The story has been corrected to say that Netflix gets 80% of its traffic from its recommendation engine.