日本語
Chatbots, step aside.
When it comes to telecom, AI is out to prove that it isn’t all talk. Early deployments of AI and machine learning (ML) reveal progress on this front, with demonstrated capabilities to transform network operations processes.
At Spirent, we gain unique insights into trends and momentum through our global work with the major mobile ecosystem stakeholders. From this vantage point, we are already seeing implementation of high-priority AI-related use cases like traffic prediction to optimize network design and investment, fault prediction to identify potential issues before they impact customers, and advancements in security posture through threat prediction and fraud detection.
We have been especially impressed by substantial environmental and cost efficiency gains that AI and ML are supporting in the radio access network (RAN) via power utilization optimization.
The combination of analytics, automation, and AI has potential to move telecom from human-intensive, rules-based processes to environments powered by machine-based decisions with human assistance, and ultimately machine-driven autonomy. At the heart of this evolution is unprecedented AI processing capabilities combined with the growth and accessibility of real-time and historical data to improve predictive modelling.
The complementary way generative AI (GenAI) can support content creation and querying is changing how telecom troubleshoots, reports, and configures complex networks. Its ability to support predictive models is enhancing accuracy and supercharging use cases, such as anomaly detection through identifying patterns that indicate network issues or security threats. Instead of looking for a needle in a haystack, AI-driven root cause analysis is accurate, helps proactively resolves issues, and frees up engineering time.
Following are the primary predictions of AI’s impact on telco networks that I shared during a recent RCR Telco AI Forum keynote.
How AI Will Impact Telco Networks
AI could enhance telco networks on numerous fronts. There are four initial areas where capital investment and the opportunity for high value returns are likely to be seen:
Radio networks reduce data flows and energy. Between 70% and 80% of an operator's total energy consumption is attributed to the radio access network. In the short to medium term, there will be continued focus on AI/ML optimizations, both autonomously in the Radio Unit, but also coordinated through the new Intelligent Controller, such as link adaption, sleep modes, and low energy schedulers. Longer term, a new form of compression using GenAI semantic communications might send just 5% of a message and accurately regenerate 100% at the other end, thus saving energy through reduction of raw data transmissions.
Devices offer more powerful features. We have already been introduced to the first wave of smartphones offering AI-enabled features for services like language translation and photo editing. This will explode with on-device GenAI multi-modal large language models powered by novel energy-efficient neural processing units (NPUs) built onto the chip. Longer term, expect on-device AI distributed processing with cloud/edge AI for a more powerful, secure and feature-rich hybrid mode.
Data centers harness cost-efficient Ethernet for AI growth. The exponential growth and rearchitecting of data centers to support hundreds of thousands of GPUs is triggering demand for cost-efficient Ethernet technology for the backend AI interconnect fabric. Soon, data center providers will start adopting the RoCEv2 protocol that enables Remote Direct Memory Access (RDMA) over an Ethernet network, while upgrading networks from 400G to 800G. Longer term, there will be adoption of a new Ultra Ethernet Transport (UET) specification, which is dedicated to AI and high-performance computing networks. UET is the fledgling work of the new Ultra Ethernet Consortium of leading experts.
Network operations become automated. Use cases with clear business value are already being adopted by telcos and will increase as automation permeates the network and AIOps enters its lifecycle, resulting in varying and decreasing degrees of human supervision. Longer term, probably in the next decade, expect full autonomy governed and enabled by network digital twins that enable continuous testing, safe and secure reinforcement learning, and most importantly AI transparency and understandability.
The Telco Path Forward
The combination of analytics, automation, and AI offers benefits to the network now, and even more amazing opportunities in the future.
AI’s utilization by telcos has been cautious and narrowly focused but this will change as experience cultivates confidence. For the network, automation of the lifecycle is essential to drive the strategic view and should form the basis for business cases. Once automation is in place, it can be enhanced with AI.
As an industry leader in testing, automation, and assurance for mission-critical industries, Spirent recognizes organizations can safely unlock AI’s full potential and transform marginal gains into monumental results by quantifying use cases, developing a data architecture and management strategy, pursuing automation, ensuring efficacy and testing security.
For more insights on AI in telecom, watch the RCR Telco AI Forum keynote presentation on-demand and read the key findings report for additional details.
