Forget performance — the only three goals for AI-native 6G systems that matter

By Catherine Sbeglia Nin October 7, 2024

Collected at: https://www.rcrwireless.com/20241007/fundamentals/forget-performance-the-only-three-goals-for-ai-native-6g-systems-that-matter

For Rakuten Mobile’s SVP of Next Generation Advanced Research, the key is to decouple 6G workloads from the cloud-native platform

In a recent article, RCR Wireless News laid out the definition of AI-native as it pertains to 6G; however, in an even more recent conversation with Rakuten Mobile’s Senior Vice President of Next Generation Advanced Research David Soldani, that definition was challenged, or at least, expanded upon.

Soldani’s main critique of current ideations around AI-native is that it’s been mostly focused on enhancing network infrastructure with AI capabilities to improve performance. “But I don’t care about performance,” he told RCR. “We already have excellent speed and latency… It’s very difficult to find new use cases enabled by the current view of how AI should be used within next gen systems.”

Instead, he offered: “AI-native will mean that the workload has the ability to be integrated seamlessly through APIs to my AI capabilities.” He added that the conversation must shift towards evolving our 5G/5G-A workloads in the direction of achieving three key goal indicators: improve revenues; reduce costs; and reduce risks. Anything else, he said, isn’t really of any concern to him.

To achieve Soldani’s vision of 6G there must be a “fundamental shift from virtual network functions and we need “100% cloud-native functions” that are in compliance with the requirements from the Cloud Native Computing Foundation (CNCF). “The CNCF should define how 6G has to be designed and then from there, we need to absolutely decouple the functions — the workloads — from my platform. It has to be Kubernetes and evolution based. That’s it,” he said. “The key is the decoupling of 6G workloads from the cloud-native platform and offloading key functionalities — such as networking, observability, tracing and security — to the platform.”

Soldani also explained that telcos don’t want vendors to offer them embedded AI algorithms for the RAN, and they certainly don’t want to pay for them. “I have my own capabilities that I want to be able to use with your functionalities,” he said.

Kailem Anderson, the Vice President of OSS/BSS provider Blue Planet — a Ciena company — offered a more comprehensive explanation of Soldani’s perspective here: “Today, the RAN vendors build their own AI capabilities into the software that enables the RAN and it’s closed, meaning that telcos rely on the vendor to figure it out. He added, however, that in a 6G environment, he expects a “democratization of AI to play out” across the entire network. “Instead of being beholden to the vendors that create the algorithms… telcos will be able to bring their own AI and apply it to the data sets that are being leveraged within the RAN environment,” he continued.

According to Anderson, this democratization will create a new power dynamic that will enable the rapid acceleration of AI within the RAN, which will support brand new use cases, and most importantly, new monetization opportunities. “Being able to leverage that data for customer care systems, or to drive value-added automation, maybe even automation that is inter-carrier — from carrier-to-carrier — because you’re not innovating just at the pace that the vendor is providing their AI into their equipment, you can accelerate on top of that,” he said.