Telecom was one of the earliest users of AI, long before its marketing hype and even before it was known by that name. The recent advancements are further propelling AI's use.

In this episode, I talk to Dan Warren, Director of Communications Research at Samsung, regarding the role of AI in telecom networks. We discuss its role in network operations, AI for RAN, AI with RAN, and AI on RAN concepts; how software-based networking and virtualization (vRAN/Open RAN) enable AI; who will develop and implement AI models; the scope of standardization; and more.

We also delve into the investment challenges for 5G operators to leverage AI, whether AI will encourage a larger role for Hyperscalers in telecom, and whether large-scale AI implementation can start with 5G/5G Advanced or will have to wait for 6G.

Highlights:

00:00 - Intro
02:10 - Guest intro (Dan Warren)
03:22 - Samsung Network's categorization of the role of AI - Networks for AI, AI for networks
04:44 - Current interesting use cases of AI - RAN energy efficiency, complex capacity, and coverage optimization
06:40 - The current status of use of AI in telecom, major industry focus areas - optimizing opex and capex, experience enhancement
90:15 - Critical role of software-based networking and virtualization in enabling AI
12:48 - Are legacy networks w/o software-based networks out of luck for AI?
15:40 - How to decide where to run AI workload - where is the AI compute needed?
19:05 - Who will develop AI models for telecom? - property vs. standard, differentiation etc.
24:35 - Dichotomy between AI differentiation, open networking, and offering AI as a service layer
26:55 - Samsung's approach to AI as a service/software leveraging its software-based networking legacy. Might be different for more established players?
29:06 - Does today's architecture (e.g., SMO) have hooks for managing AI end-to-end?
30:01 - Data challenge of operators for AI - different forms, formats, sources, granularity, etc. Will things like NWDAF solve it?
31:50 - Operator understanding of the AI challenges ahead - transforming from a hardware operator to a software management company
34:36 - Is the investment needed for AI an impediment to its adoption? Worries of the fast-moving and changing AI landscape
38:46 - AI on RAN - Samsung Network's views
40:43 - AI on RAN - operators moving from telecom service providers to AI (edge) Data Center/infra providers - Does it make sense?
43:19 - AI timing - Will large-scale deployments wait for 6G, or can start with 5G?
45:16 - Closing

The mind-numbing investment in AI and the sky-high valuations of the companies involved have led many to wonder whether AI has become a bubble.

In this episode, I talk to Dr. Riad Hartani, who is an advisor, investor, and a longtime time industry participant, on what is driving the investment - real market demand or FOMO, who is making those them, ongoing circular investment and risks, especially in Neoclouds, RoI concerns, consumer vs. enterprise AI dynamics, whether AI is a bubble, what could be early indicators of a burst, if any, who will be worst affected, and more.

We also compare and contrast the current AI situation with the Dot-com bubble.

Qualcomm surprised the IoT market by announcing the acquisition of the open-source hardware champion, with a huge hobbyist/enthusiast builder ecosystem.

In this episode, I talk to Qualcomm's VP of Product Management, Industrial and IoT BU, Manvinder (Manny) Singh, about the reasoning for the acquisition, challenges of the IoT industry, how AI is changing the game, what Qualcomm brings to the ecosystem, how AI Hub and the other acquisition, Edge Impulse, fit together with Arduino, and more.

We also delve into what will and will not change for the Arduino's open-source spirit, as well as the ecosystem, which primarily utilizes microcontrollers and low-power processors.

Also, check out Episode 59- Discussion with CEO of Edge AI Foundation:

https://bit.ly/TantrasMantra_ep59