AI with a Human Touch - The Future of Telecoms
March 2024
Microsoft presentation Building the New AI Economy at MWC24. Credit: E. Piciocchi
INTRODUCTION
Artificial Intelligence (AI) was at the heart of the 2024 Mobile World Congress (MWC24), with Generative AI (GenAI) a buzzword and major players like Nvidia showcasing powerful AI chips. This trend aligned with the growing importance of edge computing for real-time AI inference. Cloud companies were also prominent, highlighting the critical role of infrastructure for these advancements. Operators addressed how to optimise processes and potentially diversify into AI-related infrastructure services. Partnerships for GenAI and Large Language Model (LLM) development were also discussed.
Building on the MWC24 talks, exhibits, and announcements, this article delves into key AI takeaways from the event. We will explore industry studies on AI adoption, how leading telcos are approaching AI implementation, and the exciting potential of AI in customer service, RAN (Radio Access Networks), and edge computing. We will also examine the challenges and collaborations shaping the future of AI in telecoms.
Humanising AI was one the key themes of MWC24. Credit: GSMA
HUMANISING AI
The Rise of AI in Telecommunications: Industry Studies
Telecom industry reports highlight a surge in adoption of artificial intelligence (AI) and its potential to revolutionise the sector, with findings from Nvidia's State of AI in Telecommunications survey, a study by Ernst & Young commissioned by Liberty Global, and Accenture’s Pulse of Change Index.
Nvidia’s State of AI in Telecommunications Survey: Strong Endorsement for AI
The February 2024 survey State of AI in Telecommunications: 2023 Trends by chipmaker NVIDIA suggests a surge in interest and adoption of artificial intelligence (AI) within the telecoms industry. The survey, encompassing over 400 industry professionals globally, found strong endorsement for AI's potential to boost revenue and cut costs.
At MWC, Lilac Ilan, Head of Global Telco, AI-powered Operations, NVIDIA, provided GenAI and AI use case examples under four key workstreams:
Productivity: Chatbots, tools, knowledge summarization can make employees more productive.
Customer Experience: A well-informed AI chatbot can provide a more contextual explanation for bill changes, leading to a better customer experience.
Network Operations: AI-powered network and AI-assistance for field technicians can improve network efficiency and ultimately customer experience.
Revenue generation: use of avatars for the retail workforce (e.g. Viatel), AI inference at the edge for enterprises (especially with the rise of sovereign Clouds)
Liberty Global’s Study: Telecoms Key to Unlocking AI Efficiency Gains
A February 2024 study by Ernst & Young, commissioned by Liberty Global, highlighted the critical role of telecoms in enabling widespread AI adoption and maximising its economic benefits. Its key findings include:
50% of jobs in developed economies could benefit from AI, relying heavily on telecoms networks.
AI implementation across telecoms could unlock substantial efficiency gains, potentially saving $33 billion annually.
The US, Germany, UK, and France are expected to see the biggest benefits.
AI can transform the telecom sector by:
Enhancing network operations
Boosting security
Improving customer service
Streamlining back-office processes (e.g., using AI to predict and prevent network congestion)
Speaking at MWC24, Mike Fries, Liberty Global CEO, acknowledged the transformative potential of AI but emphasised the importance of adopting AI in an “organised and thoughtful way”. “Empathy”, “integrity”, and outside-the-box thinking are the qualities Mr Fries seeks in the next generation of leaders.
Accenture’s Pulse of Change Index: AI Readiness Required
A recent Accenture survey The Pulse of Change Index revealed that 97% of 3,400 C-suite executives across industries believe AI will fundamentally change their core business within the next few years, yet less than a third feel prepared.
Despite seeing generative AI as an opportunity for growth (76%), nearly half of businesses (47%) feel unprepared for the accelerating pace of technological change. Only 27% believe they are currently ready to fully leverage generative AI.
A significant gap between businesses that heavily leverage technology for growth and those that don't. Only 9% of companies are considered ‘Reinventors’ are truly integrating artificial intelligence (AI) and data-driven strategies into their core operations, but this leading group is experiencing significant benefits.
During her presentation at MWC24, Laura Peterson, Senior Managing Director, Accenture, highlighted the growing importance of AI in transforming businesses, particularly customer experience. She outlined five imperatives to focus on for successful AI implementation:
Leading with Value: Move beyond siloed use cases and focus on end-to-end capabilities to drive real business change.
AI-Enabled Secure Digital Core: Develop a secure digital foundation with the right data infrastructure to power AI initiatives.
Talent Strategy: Invest in upskilling your workforce and ensure you have the right talent to work effectively with AI.
Responsible AI: Implement AI responsibly, considering both ethical and compliance aspects.
Continuous Reinvention: Build the capability to adapt and evolve constantly.
During the keynote The Digital Vision for Telcos, Julie Sweet, Chair & CEO, Accenture highlighted several key challenges that can hinder AI adoption in the telecom sector:
Legacy systems. The network infrastructure will play a crucial role in supporting advancements in AI, particularly with edge computing.
Technical debt. Collaboration and partnerships, including with Hyperscalers, are key to overcoming these challenges.
Independent operating units and a deep engineering culture. Foster change and encourage collaboration across teams and departments by aligning incentives for successful AI adoption.
Lack of investment in HR. The success of AI implementation relies on upskilling the workforce and adapting HR strategies to a skills-based approach.
Brad Smith, Vice Chair and President, Microsoft, presents Microsoft AI Access Principles at MWC24. Photo credit: E. Piciocchi
Telcos’ approach for AI implementation
This chapter explores how leading telecommunications companies are leveraging Artificial Intelligence (AI) to transform their businesses. We begin with Deutsche Telekom's CEO, Tim Höttges, outlining key considerations for successful AI adoption, followed by in-depth analyses of specific strategies from companies like BT, Vodafone, Microsoft, and others. Each section examines their unique approaches to AI implementation, highlighting success stories and challenges encountered.
Deutsche Telekom's CEO: Embracing AI for a Telco's Future
During his presentation AI Has Come to Stay, Deutsche Telekom's CEO Tim Höttges noted that AI is revolutionising every business model and operating process, and presented his vision for how telcos can leverage AI to transform their businesses.
Here are the key takeaways:
Trust and Bias: AI models can be prone to errors and biases. Careful data selection and governance are crucial.
Human-Machine Collaboration: Finding the right balance between human expertise and AI automation is essential.
Embrace a "Shaper" and “Facilitator” Mentality: Go beyond simply using off-the-shelf solutions and fine-tune existing models, and thereafter serve the SME market.
Multi-LLM Strategy: Don't rely on a single Large Language Model (LLM). Explore different models and consider creating your own for telecom-specific needs (Deutsche Telekom is collaborating with other operators on this).
Centralised AI Competence Centre: Establish a central body to manage AI adoption across the organization, ensuring ethical standards, data security, and knowledge sharing.
CEO involvement: Lead the AI transformation from the top. Define a clear AI strategy for your organization.
Skills and Culture: Address employee concerns and educate them about the benefits of AI. Invest in building AI expertise within your company.
Hoettges also highlighted some of the 400 AI use cases already implemented within the company, such as:
Customer Experience:
AI-enhanced Chatbots with improved conversation flow and resolution rates (50% increase in first-call resolution).
Fibre rollout:
AI Mapping for FTTH Buildout and Planning (75% increase in productivity)
AI-powered Chatbot for Field Technicians (€1.5 million saved within the first six months of implementation)
Network Optimisation:
Predictive Maintenance and Anomaly Detection.
Automated Network Steering for Energy Savings (16 GWh saved in Germany).
Business Process Efficiency:
Legacy Code Translation using AI tools.
Personal AI Assistant for Employees (internal tool called "Compass").
Cybersecurity:
AI-powered threat detection and malware prevention (prevented 15,000 unauthorised access attempts).
Höttges also showcased T-phone, an app-free AI-powered phone concept developed with US partners Brain.ai and Qualcomm, suggesting a potential future where apps are obsolete.
BT’s AI for Business Transformation
BT has created over £150m in value by December 2023 through AI and data initiatives, such as increased productivity. Speaking at the "Can Telcos Afford Not to be Part of the AI Race" session at MWC, Harmeen Mehta, Chief Digital & Innovation Officer at BT Group, shared her insights on BT's successful implementation of AI.
Here are the key takeaways:
Data Strategy: Consolidated and cleansed a vast amount of data
Focus on Use Cases: Identified key business problems to solve with AI, not just technological advancement
Balancing Technology and Humanity: Emphasised the importance of human expertise alongside AI capabilities.
Mitigating Bias: Acknowledged the risk of bias in AI models and actively works to address it.
Flexibility: Established an "LLM gateway" for flexible algorithm switching to handle future uncertainties.
In conclusion, Ms Mehta stated that “the power of AI together with the power of human will create the super human of tomorrow”.
Vodafone’s AI Implementation Approach
Speaking at the "Can Telcos Afford Not to be Part of the AI Race" session at MWC, Scott Petty, CTO of Vodafone Group, shared Vodafone's key considerations for a successful AI implementation:
Business-Driven Approach: AI should be viewed as a business enabler, not just a technology project.
Balancing Human and Machine: Finding the right balance between human oversight and AI automation based on factors like data sensitivity and risk tolerance.
Understanding Different LLMs (Large Language Models): LLMs have varying capabilities and cost structures, requiring careful selection for specific use cases.
Pettey also discussed three Categories of Generative AI Use Cases:
Productivity: Automating tasks like legal contract analysis.
Function Transformation: Transforming customer care with "super agents" powered by generative AI.
New Products & Revenue Streams: Vodafone is exploring innovative generative AI applications that they cannot disclose due to competition.
Telstra's AI Strategy
At MWC, Telstra CEO Vicki Brady emphasised collaboration (e.g., Accenture, Microsoft) and internal cultural shift. Telstra uses AI to address business challenges and improve customer service (e.g., Ask Telstra digital assistant). Ms. Brady sees AI as a company-wide strategy, not just a technological one.
Ethio Telecom’s investment in AI Talent
During the keynote The Digital Vision for Telcos, Ethio Telecom CEO, Frehiwot Tamiru, stressed the importance of developing talent with AI skills, and stated that talent strategy is seen as equally important as technology and marketing for industry transformation. Collaboration with educational institutions is key to incorporating AI into curriculums.
Amdocs’ approach to GenAI
Anthony Goonetilleke, group president of technology and head of strategy at Amdocs, described the potential and challenges of GenAI in the communications industry.
Generative AI will revolutionise telecoms, by transforming network operations and customer interactions. Those who don't adopt it risk falling behind.
However, success hinges on reliable service, consistency in pricing and responses to inquiries, trustworthy AI, and high accuracy.
Organisations need to adapt talent, data, cybersecurity, tools, governance and goals to embrace AI. In particular, Amdocs:
Defined a clear data strategy and ingestion pyramid, governance approach, architecture, and railguards.
Fine-tuned GenAI co-pilots to augment human capabilities in various domains and improve accuracy and optimise costs.
Built a telco-specific Generative AI platform through collaboration with industry leaders, focusing on innovation, responsibility, and cost-efficiency
Prioritised business-led initiatives with ROI models and achieved early results with Amdocs' amAIz platform.
Telefónica: Beyond Automation
Elena Gil, AI & Big Data Director at Telefónica Tech, chaired their "Beyond Automation" session, showcasing Telefónica's internal AI projects and their impact, as well as their end-to-end AI value proposition for both public and private customers.
According to Ms. Gil, Telefónica :
Has extensive experience leveraging AI (both traditional AI and GenAI) to optimise its operations (network efficiency) and customer engagement - Text generation, data analysis assistance, and internal information search are initial areas of focus. Multimodal applications are also being explored - .
Offers a suite of advanced digital technologies to boost efficiency, sustainability, and resilience - in diverse sectors like energy, healthcare, and fitness - .
Integrates GenAI with cybersecurity, cloud, IoT, big data, and blockchain to create innovative solutions for B2B clients. These solutions can automate tasks, optimise processes, and improve decision-making.
Ms. Gil also emphasised the importance of responsible AI development, aligning with UNESCO's principles on ethical AI use and human rights.
Nokia’s vision for the Future Network
During his presentation Beyond ChatGPT: The Future of Generative AI in Business, Jitin Bhandari, Nokia's CTO & VP of Cloud & Network Services, presented a vision for the future of communication networks driven by Generative AI.
Bhandari emphasised five key principles for building this future network:
Expertise
AI Governance
Machine Learning Technologies
Data Frameworks
AI Applications (developer communities)
While acknowledging the excitement surrounding ChatGPT, Bhandari proposed a dual approach of Large Language Models (LLMs) and Traditional AI and times series functions, which will continue to be relevant for network monitoring and analysis.
Bhandari described the building blocks of Nokia's Telco AI a/ GenAI platform, designed to tackle both productivity and product use cases in the telecom sector:
Model Farm: A combination of open and closed LLMs.
Telco Vectorisation: Making generic LLMs understand telco-specific data and language.
Domain Knowledge / Fine Tuning / Prompt Engineering / Hallucination management: Incorporating knowledge from past projects and fine-tuning prompts to improve accuracy, guide LLMs and reduce irrelevant outputs.
Microsoft AI ecosystem
In his presentation titled The Emerging AI Ecosystem, Microsoft Vice Chair & President Brad Smith compared the potential impact of artificial intelligence (AI) on reshaping education and the global economy to the revolutionary effects of the printing press. Smith described Microsoft approach in the AI ecosystem, which involves extensive collaboration, infrastructure investment, openness, and respect for user data privacy and choice:
Microsoft supports the development of technology in Europe. At MWC, Microsoft announced USD 5bn of investments in data centres (DCs) in Europe.
Microsoft relies on partnerships. During MWC, Microsoft signed a deal with French AI startup Mistral to train Mistral’s AI data model on Microsoft’s Azure and to develop AI solutions collaboratively.
Unlike some competitors, Microsoft allows app developers to choose whether to use its marketplace.
Developers who opt to train their models with Microsoft retain ownership of their data, with Microsoft committing not to access or compete with it.
Microsoft gives users the freedom to choose their cloud service provider.
Leaders from Amdocs, Amelia, 5G Digital Catapult, and NVIDIA discuss The Customer Experience Revolution in the AI Era at MWC24. Photo credit: E. Piciocchi
AI for Customer Experience
This chapter dives into the transformative power of Generative Artificial Intelligence (GenAI) in customer service. We explore a panel discussion featuring industry leaders from NVIDIA, Amdocs, Amelia, and 5G Digital Catapult, who delve into the benefits, challenges, and recommendations for successful AI implementation. Additionally, we examine the efforts of the Global Telco AI Alliance (GTAA) and the partnership between ServiceNow and NVIDIA to develop AI solutions specifically tailored for the telecom industry.
Panel - The Customer Experience Revolution in the AI Era
In a discussion titled "The Customer Experience Revolution in the AI Era," a panel of industry leaders explored the potential of Generative AI (GenAI) to transform customer service.
Lilac Ilan, Head of Global Telco, AI powered Operations, NVIDIA
Gadi Solotorevsky, Head of Data Engineering, (G)AI and ML, Amdocs Data Intelligence, Amdocs
Chetan Dube, CEO, Amelia
Dritan Kaleshi, Director, Technology, 5G, Digital Catapult
They delved into the benefits, challenges, and recommendations for successful AI implementation, along with predictions for the future evolution of customer experience.
Benefits
The panellists discussed the benefits of GenAI in customer experience:
Anticipation: with the introduction of GenAI, there's a shift towards understanding not just what customers have done but also what they want, potentially revolutionising customer service and satisfaction. (Solotorevsky)
Hyper-personalisation: AI knows everything about a customer, such as their preferences and buying habits, and can help you personalise your offers. (Dube)
Smooth customer journey: Identify and solve pain points in customer journey, starting with purchase, by making it more personalised, dynamic and proactive through the use of an AI-powered avatar. (Ilan)
Challenges
The panellists described several critical factors that AI implementation faces:
Limitations of AI Technology: LLMs can make mistakes. We need to consider what happens if they provide incorrect explanations or offer non-existent products. (Solotorevsky)
Substandard AI-generated responses: Air Canada recently lost a lawsuit and is liable for a customer's financial loss caused by misinformation provided by a chatbot on the airline's website. This underlines the risks of AI implementation in customer service. (Ilan)
Messy Data: "Garbage in, garbage out" applies to AI. While big data isn't always necessary, data needs to be accurate and relevant for AI to function effectively. (Solotorevsky)
Incredible space of development: We're witnessing the largest human experiment with AI ever and we are going four times faster every six months. (Kaleshi)
Mission critical tasks: Harder to incorporate GenAI tools in mission critical tasks than to use them for simple task productivity. (Kaleshi)
Digital workforce management: Executives are concerned about the definition of KPIs and OKRs of their digital employees, and the blue print on how to get there. (Dube)
Inability to quantify ROI: 44% of the participants Nvidia survey finds the inability to quantify ROI the biggest challenge to achieving AI goals. (Ilan)
Recommendations:
The panellists provided the following advice:
Use case prioritisation: Get the business objectives first, define KPIs, priotise use cases, e.g. time to market. (Ilan)
Focus on Both Human and Digital Employees: A hybrid workforce is the future, and we need to establish frameworks for how humans and digital employees will work together. (Dube)
Data Strategy: The success of AI relies heavily on data quality. Curate the right data and define your strategy for handling data to provide the right information for AI generation. (Ilan)
Accelerate compute: Develop the infrastructure you will need to make the AI grip.
Alignment with Business Goals: AI technology needs to be aligned with your business goals and commitment to responsible AI. (Ilan)
Skilling the Workforce: Upskilling the workforce to understand and work effectively with AI is crucial, put technology expertise and domain expertise together. (Kaleshi)
Testing and Validation: Frameworks and methodologies for testing and validating AI outputs are critical for ensuring responsible and reliable AI, and must come from the domain expertise. (Kaleshi)
Responsible AI: AI development and deployment need to be aligned with responsible AI practices. (Ilan)
Looking Ahead: Future Evolution of Customer Service
The next five years will likely see significant advancements in AI, including:
Native AI Platforms: The development of AI platforms specifically designed for ease of use and integration – instead of a silo approach – . (Dube)
Fewer hallucinations: The reduction in hallucinations by increasing the size of data sets increasing and by laying cognitive determinism on top of generative framework in order to offer enterprise grade services. (Dube)
Improved Testing Tools: The creation of more robust tools for testing and validating AI outputs. (Kaleshi)
Dramatic Customer Experience Improvements: We can expect a significant jump in customer experience due to AI adoption, provided it's accompanied by proper verification practices. (Kaleshi)
KPI improvements: The use of digital workforce to reduce customer service cost, improve Net Promoter Score, reduce customer churn, increase cross-sell and upsell. (Dube)
Customer evolution: The evolving role of GenAI in customer service - customers will have their own GenAI assistants -. (Solotorevsky)
Multimodal Customer Interactions: The way customers interact with AI will evolve beyond text-based interactions to include video and other modalities. (Ilan)
Chetan Dube concluded by stating that Telcos that use AI will get ahead.
GTAA will create LLM for AI-powered Customer Service
SK Telecom, e&, Singtel, Deutsche Telekom, and SoftBank are creating the Global Telco AI Alliance (GTAA), a joint venture focused on Large Language Model (LLM) development.
Goal: Develop AI models tailored for telecom companies' needs, initially targeting their combined 1.3 billion customers across 50 countries.
Focus: Personalised customer service through AI-powered digital assistants.
Multilingual Support: Initial focus on languages spoken in partner markets, with plans to expand later.
Collaboration with Experts: The venture is working with researchers and language specialists to ensure accurate representation of less-common languages.
Development Underway: Work on the telecom-specific LLMs has already begun.
GTAA signifies a significant step towards AI-powered customer service in the telecom industry. By leveraging combined resources and expertise, the alliance aims to create advanced AI models that cater to the specific needs of global telecom companies and their customers.
ServiceNow and NVIDIA Partner on Telco-Specific AI Solutions
ServiceNow and NVIDIA are expanding their partnership to develop AI solutions specifically for the telecom industry.
The first offering, Now Assist for Telecommunications Service Management (TSM), leverages NVIDIA AI technology to improve customer service experiences. Built on the Now Platform, Now Assist aims to:
Boost agent productivity: Through AI-powered features like chat summarization and agent assistance.
Speed up issue resolution: By helping agents understand complex technical problems faster.
Enhance customer experience: By providing efficient and accurate support.
This partnership is a response to the growing demand for AI in the telecom sector, with 73% of global service providers prioritizing AI investments. ServiceNow and NVIDIA believe their collaboration will create significant business value for telcos. ServiceNow and NVIDIA plan to develop further AI solutions tailored to the specific needs of the telecom industry.
DeepMind CEO Highlights the Power of Artificial Intelligence
In a conversation with Steven Levy, Editor at Large, Wired, Demis Hassabis, Co-Founder & CEO, Google DeepMind, explored the potential of Artificial General Intelligence (AGI) and its real-world applications. He also addressed the recent controversies surrounding Google's the image generation of its generative AI (genAI) tool, Gemini.
From Games to General Intelligence: The video game industry was a breeding ground for advanced AI techniques.
Unlocking Potential, Defining Success: AGI is envisioned as a problem-solving powerhouse, capable of tackling complex challenges that elude human capabilities.
AI as a Scientific Catalyst: DeepMind's AlphaFold program demonstrates the potential of AI to unlock fundamental scientific questions, such as protein folding. It predicts protein structures crucial for medicine and drug discovery, significantly accelerating research.
Impact Across Fields: Similar progress is expected in materials science, weather prediction, and mathematics.
Consumer adoption: The success of Google’s competitor OpenAI showed consumer enthusiasm for GenAI, despite its imperfections.
Business and Innovation: The merger of DeepMind with the AI research unit Google Brain led to a focus on product development like the recently launched Gemini model.
Gemini’s image generation controversy: The feature intended to be universally appealing but lacked nuance and generated inappropriate images, particularly for historical figures. Google temporarily disabled and is refining the technology and addressing these issues.
The Ethics of Superintelligence: AGI could fall into the wrong hands. Robust ethical guidelines should govern its development and deployment.
The Future of Human-Computer Interaction: Phones may not be the ideal future interface for AI. More immersive devices like glasses could provide AI with better context and enhance human-computer interaction.
The new AI-RAN Alliance aims at accelerating AI adoption for/and/on Radio Access Network. Photo credit: Softbank
RAN AI
Fuelled by Nvidia's stock performance and industry enthusiasm, AI emerged as a major driver for Radio Access Network (RAN) advancements at MWC24, including performance improvement, cost reduction and RAN applications.
This chapter explores the growing importance of Artificial Intelligence (AI) in Radio Access Networks (RAN), including the AI-RAN Alliance, partnerships, and development kit advancements aimed at improving performance, reducing costs, and creating new applications.
AI-RAN Alliance
On the first day of MWC24, a new AI-RAN Alliance was launched to accelerate AI adoption in RAN. The alliance is formed by industry leaders, including:
Chip makers: Nvidia, Arm,
RAN equipment manufacturers: Nokia, Samsung, Ericsson,
Hyperscalers: AWS, Microsoft,
Mobile operators: T-Mobile, SoftBank,
Software companies: DeepSig
Academic Institutions: Northeastern University.
This collaboration focuses on improving efficiency, reducing power consumption, and unlocking new revenue opportunities through AI-powered services. Potential benefits include:
Reduced latency through edge-based AI processing.
Improved efficiency through resource allocation and spectrum usage within the RAN.
Development of innovative services like real-time robot interaction.
The Alliance members will share their data and expertise, and leverage their collective leadership. They will focus on three main areas of research and innovation:
AI for RAN – advancing RAN capabilities through AI to improve spectral efficiency.
AI and RAN – integrating AI and RAN processes to utilise infrastructure more effectively and generate new AI-driven revenue opportunities.
AI on RAN – deploying AI services at the network edge through RAN to increase operational efficiency and offer new services to mobile users.
Nokia and Nvidia Partner on AI-Powered Cloud RAN
Ahead of MWC24, Nokia announced a collaboration with NVIDIA to develop next-generation mobile networks using AI and cloud technology. This partnership aims to make Cloud RAN a reality, offering operators and businesses:
Improved Efficiency and Performance, through a combination of cutting-edge technology from both companies, including NVIDIA's Grace CPU and GPUs alongside Nokia's In-Line accelerator.
Lower Energy Consumption, through Nokia's In-Line technology designed for low power usage.
More Flexibility and Choice, through open and flexible Cloud RAN solutions, building on Nokia's anyRAN approach.
This collaboration paves the way for the commercialization of a more efficient and customizable Cloud RAN solution powered by AI and cloud technology.
Intel's vRAN AI Development Kit
Intel has released releasing the vRAN AI Development Kit to select partners, enabling them to integrate AI capabilities into virtual Radio Access Networks (vRAN) without extensive AI expertise.
The key Features of the Kit include:
Pre-built AI models: Optimized for vRAN use cases and built on industry-standard libraries and frameworks like oneAPI, TensorFlow, PyTorch, and OpenVINO.
Optimised for Intel Xeon Processors: The kit leverages built-in AI acceleration, telemetry, and power management features of Xeon processors for efficient network operation.
Customization and Integration: Training code allows partners to tailor AI solutions to specific network needs.
Reference Architectures: Show how to integrate AI models with other RAN applications and components.
The kit brings the following benefits to the operators:
Dynamic Network Reconfiguration: Intel's solution allows operators to adjust their networks based on real-time data, potentially leading to cost savings and improved revenue opportunities.
Maximised Infrastructure Value: AI-powered insights can help operators extract more value from their existing network infrastructure.
Interviewed at MWC24, Sachin Katti, SVP & GM, Intel, stated that AI plays a vital role in optimising performance and managing software-defined radio access networks (RAN). Intel's vRAN AI Development Kit allows building and deploying AI models for vRAN use cases, enabling network reconfiguration and cost savings.
Michael Dell, Chairman & CEO, Dell Technologies, highlighted the transformative role of AI and edge computing at MWC24. Photo credit: E. Piciocchi
AI at the Edge
This chapter delves into of the convergence of AI with 5G at the edge. At MWC24, leaders from Dell, Google, Amdocs, Equinix, and ZTE highlighted the transformative power of this combo, unlocking new possibilities across industries, from AI-powered manufacturing lines to real-time data analysis for autonomous vehicles. Challenges like data security and monetization models remain, but the focus on sustainability and customer benefits paints a promising future. Additionally, we examine the AI platforms from Intel and Nokia, and Qualcomm.
Dell and the Digital Imperative
During his keynote New Strategies for a New Era - The Digital Imperative, Michael Dell, Chairman & CEO, Dell Technologies, highlighted the significance of data in the telecommunications industry, and the transformative role of AI and edge computing:
Telcos as Key Players: Telcos are positioned as critical players in this data-driven landscape, having invested significantly in AI (GenAI) compared to other industries. They have the potential to facilitate transformative changes for their clients' businesses.
Edge Computing's Importance: The value of data is emphasised as being created at the edge, necessitating the movement of cloud, applications, and AI closer to where data is generated in the real world.
Utilisation of Telco Data: Telcos possess vast amounts of data, which they can leverage to train AI models and redefine various aspects of their operations, including customer support, operations, and sales.
Michael Dell also described Dell’s initiatives in the Telecom industry:
Dell allocates resources to the Telco sector, including the addition of personnel dedicated to Telco units and strategic partnerships with software companies like RedHat and Windriver.
Dell has partnerships and initiatives with major operators like AT&T and Ericsson, aimed at enhancing network openness, agility, and efficiency through virtualisation and cloud principles.
Dell aims at collaborating with the entire Telco ecosystem, facilitating virtualisation initiatives globally, whether in brownfield or greenfield environments.
Dell positions itself as a facilitator of transformation for Telcos and their customers, leveraging its expertise in technology and innovation to drive change and evolution.
Dell focuses on sustainability, aiming to reduce environmental footprint and achieve net-zero emissions, aligning with broader industry and societal goals.
Amdocs, 5G and AI at the Edge: A Game Changer Across Industries
At MWC24, Anthony Goonetilleke, Group President, Technology & Head of Strategy, Amdocs, emphasised the transformative power of 5G and AI working together at the network edge. According to Goonetilleke, the combination of 5G and AI will unlock new possibilities in various industries, including:
Enhanced AR/VR experiences with ultra-low latency,
Smart city development,
Autonomous vehicles and automated manufacturing,
Real-time, on-demand access to powerful AI applications, and
AI-powered personalised experiences in retail and logistics.
Whilst the widespread adoption of these technologies will lead to significant growth for telecom companies, challenges encompass:
Infrastructure investment for edge computing,
Regulatory considerations,
Interoperability issues, and
Data security and AI algorithm trust.
Goonetilleke recommended Telecom operators to:
Invest in infrastructure and robust ecosystems,
Develop innovative solutions for enterprises at the edge, and
Prioritise data security and trustworthy AI algorithms.
Intel's Edge Platform
Launching this quarter, Intel's Edge Platform (formerly Project Strata) will simplify deploying AI at the edge. Unlike cloud platforms, it is built for distributed edge environments with existing infrastructure in mind. Targets include defect detection in manufacturing, frictionless checkout in retail, and traffic management in smart cities. Launch partners will include Amazon Web Services, Lenovo, L&T Technology Services, Red Hat, SAP, Vericast, Verizon Business and Wipro.
Interviewed at MWC24, Sachin Katti, SVP & GM, Intel, emphasises their broad portfolio across software, silicon, and platforms to enable AI across the cloud, network, edge, and devices. Katti described the impact of Edge AI across industries. Enterprises use edge AI to improve competitiveness, automate operations, and innovate faster. Applications include AI-powered product recommendations in retail, AI analysis of athletes in sports, and GenAI assistants for order taking in restaurants.
Nokia’s MX Workmate, an AI Platform to Empower Industrial Workers
Ahead of MWC24, Nokia launched MX Workmate, a platform that adapts Generative AI (GenAI) technology for operational technology (OT) environments, with the following key features:
Tailored for Industrial Use: MX Workmate addresses the specific needs of industrial settings, unlike generic GenAI models.
Real-Time Contextual Information: Provides workers with relevant and up-to-date information in natural language.
Security and Robustness: Designed for mission-critical applications with enhanced security features.
Bridging the Skills Gap: Aims to compensate for the lack of specialised skills in the workforce.
MX Workmate aims at providing businesses with the following benefits:
Improved Efficiency and Productivity: Equips workers with real-time information for better decision-making.
Enhanced IT/OT Integration: Bridges the gap between Information Technology (IT) and Operational Technology (OT) operations.
Digitalization Strategy: Supports companies in their digitalization efforts by addressing workforce expertise challenges.
Qualcomm’s AI Hub
Qualcomm intends to replicate Nvidia's success in AI training with AI Hub, a platform for deploying generative AI services at the edge. AI Hub offers a remote test bed for apps using AI models and supports various devices, with an emphasis on Qualcomm silicon for optimised performance. Qualcomm aims to establish AI Hub as the go-to platform for edge-based generative AI, potentially increasing market share and chip profitability.
Further, Qualcomm unveiled the Snapdragon X80 5G modem-RF system featuring a dedicated 5G AI processor and support for six-carrier aggregation.
Google's GDC edge
During the Convergence of AI and 5G at the Edge session, Ankur Jain, VP Telecom Industry & Google Distributed Cloud (GDC) at Google, presented Google's perspective on edge computing.
Google's definition of the Edge encompasses a range from a couple of servers to potentially hundreds, on customer premises. Notably, this does not include devices such as smartphones and IoT devices, nor public Cloud data centres.
Over the past few years, Google have seen businesses leveraging edge and 5G to transform their businesses, increasingly so with the emergence of AI capabilities at the edge.
Customers want a common platform the ability to select and deploy applications of their choice at the Edge, preserving the scalability and operability of the public Cloud, as well as safeguarding sovereignty, privacy and security. Total Cost of Ownership (TCO), inclusive of the operations maintenance costs of edge deployments, also plays a significant role in decision-making.
GDC as a fully managed cloud in on-premise Cloud infrastructure platform plus a Managed Services platform, where on top of it Google is working with the entire ecosystem, including the application developers that are building solutions for that particular industry, or the delivery partners who are helping Google get this deployed on the customer premises
Google addresses three main Edge use cases:
Data Localisation: Some data necessitates remaining on-premises due to compliance or other considerations. A hybrid approach, integrating on-premises infrastructure with cloud services, ensures a seamless user experience. Training occurs on the cloud, while inference takes place at the Edge.
Network Evolution: Edge computing proves highly efficient for network-intensive tasks, particularly when located close to the source of traffic, enhancing overall network performance.
Modern Retail Stores: Jain highlighted the relevance of Edge computing in retail stores and quick service restaurants but also on manufacturing floors. Clients want a platform that can continue to run the most mission practical services at the edge, such as points of sales for a retail chain that has backhaul constraints, or high bandwidth camera feeds in a private wireless network.
Panel: Convergence of 5G & AI at the Edge
At MWC24, the Convergence of 5G & AI at the Edge panel discussion shed light on the rationale and opportunities presented by the convergence of 5G and AI at the edge, emphasising the need for a customer-centric approach and monetisation to unlock the full potential of these technologies.
Wipro's Suruchi Gupta, GM & Global Head of Technology Development for 5G and Edge Connectivity, led the panel and speakers included:
Xue Tang, VP, and Vice General Manager of RAN Product, ZTE Corporation
Anthony Goonetilleke, Group President, Technology & Head of Strategy, Amdocs
Jillian Kaplan, Head of Global 5G & Telecom Thought Leadership, Dell Technologies
Jim Poole, VP, Global Business Development, Equinix
Rationale for the convergence:
The panellists discussed the driving forces behind the convergence of 5G, AI, and edge computing, emphasising factors like data volume, privacy concerns, and sustainability.
Sustainability in technology decisions is growing in importance, and significant energy savings are achieved by processing data locally instead of sending it to the Cloud. (Kaplan) (Tang)
Research by GSMA Intelligence, Dell Technologies, and Intel (The Next Generation of Operator Sustainability: Greener Edge and Open RAN, September 2023) estimates that retaining just 20% of traffic at the network edge, instead of sending it to centralized data centres, could lead to a 15% reduction in overall energy consumption. (Kaplan)
The convergence of 5G (throughput, latency) and AI (intelligence) at the edge is an inevitable trend. (Tang)
The limitations of physics, the sheer volume of data, and privacy regulations necessitate processing closer to data sources. (Poole)
Scale economics are key to meet the demands of customers across various locations and regions. Private data and public compute can coexist by separating storage and processing within regulations. (Poole)
Private networks can run edge applications built directly with clients. (Kaplan)
AI processing is crucial, particularly in predicting data flow patterns to efficiently allocate resources. (Tang)
Use cases:
Panellists also explored real world applications for 5G and AI at the edge, various industries like manufacturing, retail, healthcare, and autonomous vehicles.
5G's true potential lies in its enterprise applications, bringing AI and computing power directly to various industries like manufacturing, retail, and healthcare. (Kaplan)
5G and AI at the edge can improve worker safety, particularly in hazardous manufacturing environments. (Kaplan)
Dell has collaborated with NVIDIA and Arrow “to build AI and ComputerVision into Exhibit A Brewing Company's canning line to reduce the labour needed on the line by 25%, freeing them up to do other things that help increase revenue. This technology can be applied to any manufacturing process.” (Kaplan)
AI and 5G at the edge can be used to analyse data and optimise resource allocation in steel factories. (Tang)
The combined solution of 5G and AI can be used to send data from cars and road infrastructure to Edge computing for real-time processing and decision-making for autonomous vehicles and make autonomous driving more efficient and cost-effective. (Tang)
Video or voice data can be analysed at the edge (Poole)
Amdocs has collaborated with an open innovation lab in Seattle. Drones flew over orchards taking high-definition photos, which were uploaded to a central location for analysis, part of which was uploaded to the cloud for interpretation. Results were sent back to the farmers via a dashboard each morning, showing which areas of their crops required areas needing harvest, watering, or attention. (Goonetilleke)
Customer focus:
The speakers emphasised a customer-centric approach to edge applications, highlighting the importance of demonstrating clear business value and prioritizing user experience.
Customers care about solving problems, not the underlying technology. They want to know the impact on their business, such as saving money, increasing efficiency. (Kaplan)
Customers aren't afraid of AI when it translates to tangible benefits, such as reduced labour costs and improved processes. (Kaplan)
Focus discussions on how AI applications can improve a business's bottom line - saving money, making money, and overall improvement. (Kaplan)
Users are interested in customer experience, not technology, thus demand seamless connectivity and low latency. (Goonetilleke)
Applications need to be designed from the start to leverage the capabilities of the edge. (Goonetilleke)
Monetisation:
• Monetisation models for 5G and AI at the edge are currently in a stage of experimentation. (Goonetilleke)
• The pricing models of equipment manufacturers will evolve as AI becomes an integrated feature rather than an add-on. (Goonetilleke)
• The industry will find solutions for power, cost, and resource constraints, leading to a sustainable economic model. (Goonetilleke)
Leaders from Amdocs, Amelia, Digital Catapult, NVIDIA, and Wipro discuss the Convergence of 5G & AI at the Edge at MWC24. Photo credit: E. Piciocchi
AI Outlook
This chapter explores the hype of Artificial Intelligence at MWC24, the implications of the various operator AI initiatives and industry alliances, and charts the course for AI’s future in the telecom sector.
AI Hype at MWC24
Artificial Intelligence reigned supreme at MWC24. AI applications showcased at MWC24 included conversational interfaces for network management, wearable devices with large language model integration, and voice-controlled smartphones.
Generative AI (GenAI) is gaining traction across network management, customer service, and field operations, offering operators the potential for both revenue growth and cost reduction, especially in labour-intensive areas. For instance, businesses are leveraging GenAI's capabilities to enhance customer interactions, create marketing and sales content, and even develop computer code.
Nearly every product, service provider, and startup showcased some form of AI integration.
However, as highlighted in Xona Partners "AI in Telecom" interview at MWC, the term "AI" was often broadly applied to cover standard data analysis, mathematical optimisation, and statistical modelling techniques.
Telco initiatives
For telecom operators, two key AI initiatives emerged, one internal: AI-powered network and customer experience, and the other external: AI Enterprise Solutions.
AI-powered network and customer experience
The first initiative builds on efforts from recent years, focusing on optimising processes, networks, and operations. It also aims to enhance customer experiences, digital marketing, and monetisation strategies. Techniques like Large Language Models (LLMs) and Speech Recognition Models (SLMs) are being combined with other AI methods for prediction, grouping (clustering), and categorisation (classification) to achieve these goals. Currently, the primary focus is on cost reduction rather than revenue generation, highlighting that AI utilisation in telecoms is still evolving.
AI Enterprise Solutions
The second initiative aligns closely with ongoing efforts in Generative AI (GenAI), which aims to develop AI solutions for businesses. Operators are taking various approaches to this.
Some are partnering with chip manufacturers like Nvidia and AMD to offer Graphics Processing Unit (GPU) as a service to enterprises. Others are enhancing their AI cloud platforms with GenAI-centric applications for both consumers and businesses. The long-term development of these initiatives remains uncertain. However, it is clear that telecom operators entering this space will face competition not just from hyperscale cloud providers but also from a new wave of AI-focused cloud services (e.g., Coreweave, Lambda Labs, Cruose Cloud). These companies specialise in GPU-as-a-service for enterprises, sometimes bundled with LLMs and GenAI applications.
For some operators, the rise of AI presents an opportunity to capture a share of the AI infrastructure market. This could involve offering AI-centric Infrastructure as a Service (IaaS), Software as a Service (SaaS), or Platform as a Service (PaaS) solutions, or focusing on infrastructure specifically designed for AI inference (drawing conclusions from AI models). Training of complex AI models would likely be handled by large hyperscalers and LLM providers. However, this proposition is challenging, except potentially in vertically integrated telecom ecosystems (common in some parts of Asia) or for operators that have successfully developed an edge cloud value proposition by incorporating AI/LLM inference applications at the network edge.
Alliances
While promising alliances in AI development for telecom were announced at MWC24, a closer look reveals challenges that could stymie progress.
Global Telco AI Alliance (GTAA)
There were significant announcements at MWC regarding collaborations between telecom operators for developing GenAI and LLM technology.
A notable example is the alliance formed by Softbank, SK Telecom, Singtel, Deutsche Telekom, and e&.
While these collaborations are commendable in principle, practical implementation and long-term viability can be hindered by competition within the industry and the significant capital expenditure required.
AI-RAN Alliance
The AI-RAN Alliance could change how wireless services are deployed and bring broad innovation and operational efficiency to the telecom sector.
However, technical, profitability and strategic hurdles may come in the way.
Compatibility: Integrating Nvidia's chips with existing RAN equipment from established vendors like Ericsson and Nokia is not a straightforward process. Software adjustments and potential hardware changes will be required.
Cost: Nvidia's technology is expensive, particularly for power-hungry Layer 1 processing. Existing solutions from companies like Intel and Marvell are more economical. Mobile operators will need to assess whether potential revenue gains will outweigh the additional costs before investing in AI-powered RAN hardware.
Cybersecurity: Introducing AI to the RAN present privacy and security risks for both the AI itself and the applications in the RAN. The Alliance will need to design and implement an AI code of ethics and cybersecurity rail guards.
Control: Nvidia's leadership of the Alliance, coupled with the presence of the Cloud giants Microsoft and AWS in the alliance fuels concerns about their intent to exert greater control over network infrastructure. Thus, traditional telecom players will need to carefully evaluate the trade-offs before embracing this new approach.
The challenging road ahead
Capitalising on a robust consumer adoption in 2023, 2024 is poised to witness the initial rollout of Generative AI (GenAI) business applications, sparking an intense battle for supremacy within the AI ecosystem.
Monetising AI will prove challenging for Telcos. They will need to be pragmatic, identifying real-world applications in customer service and network efficiency, and addressing challenges like cost, privacy, cybersecurity, climate impact, and skills gaps. Ensuring a complementary relationship between the human and digital workforce will also be crucial.
AI on its own will not change the destiny of mobile operators; it is not a saving grace. Yet, AI is likely to contribute to performance improvements and cost reductions through automation and optimisation. It will also influence the RAN stack, particularly in 5.5G and 6G in the long run.
Demis Hassabis, DeepMind CEO, discusses Artificial General Intelligence (AGI) at MWC24. Photo credit: E. Piciocchi
KEY TAKEAWAYS
MWC24 was dominated by Artificial Intelligence, with Generative AI a buzzword and major players like Nvidia showcasing powerful AI chips.
The Rise of AI in Telecommunications: Industry Studies
Recent industry reports published ahead of MWC24 – Nvidia's State of AI in Telecommunications, Liberty Global's study by Ernst & Young, and Accenture's Pulse of Change Index – all underscored a surge in AI adoption and its transformative potential. These studies highlighted the potential for AI to boost revenue, cut costs, enhance security, and streamline back-office processes. However, navigating this AI transformation also presents challenges. Legacy systems, technical debt, a siloed corporate culture, and a lack of investment in upskilling the workforce were identified as significant hurdles.
Telcos’ approach for AI implementation
At MWC24, leading telecommunications companies discussed how they are approaching AI implementation. Deutsche Telekom advocated for a human-machine collaboration approach, focusing on trust, bias mitigation, and responsible AI development. BT's success story emphasised the importance of a data strategy, a focus on business problems, and balancing technology with human expertise. Vodafone highlighted the business-driven approach and the need to understand different LLMs for specific use cases.
AI for Customer Experience
MWC24 speakers described how AI could revolutionise customer service in telecom with benefits like anticipating customer needs and personalising offers. They promised a future with user-friendly AI platforms, fewer AI errors, and a dramatically improved customer experience. However, challenges like hallucinations and messy data exist. Experts recommend prioritising use cases, building a strong data strategy, and upskilling the workforce for successful AI implementation. New industry collaborations and initiatives intend to accelerate AI adoption. The Global Telco AI Alliance (GTAA) is a prime example, aiming to develop LLMs specifically designed for telecom companies' customer service needs.
RAN AI
The industry is also exploring the use of AI in Radio Access Networks (RAN). The new AI-RAN Alliance aims at fostering collaboration to improve spectral and operational efficiency, and unlock new revenue opportunities through AI-powered services, notably at the edge. Partnerships like Nokia and Nvidia's collaboration on AI-powered Cloud RAN are paving the way for a more efficient and customisable RAN powered by AI.
AI at the Edge
MWC24 speakers highlighted that the convergence of 5G's high-speed, low-latency connectivity with AI processing power at the edge unlocks a new wave of possibilities. This synergy can play a major role in industries like manufacturing, retail, healthcare, and autonomous vehicles. Successful AI implementation requires a customer-centric approach. Businesses need to demonstrate clear value propositions and prioritize seamless user experiences. Applications need to be designed from the ground up to leverage edge computing capabilities.
AI Outlook
Telcos are pursuing AI applications internally, such as optimising processes and enhancing customer experience, and externally, such as developing enterprise solutions with partners like Nvidia. However, they are facing capital expenditure limitations and competition from hyperscalers and AI-focused cloud service players. The AI-RAN Alliance will need to overcome technical compatibility and cost hurdles. Ethical, security, and environmental concerns also need to be addressed. Overall, AI will not be a silver bullet for mobile operators' future success. However, AI does hold promise for efficiency and 5.5G and 6G development.
In this article, we explored how Artificial Intelligence is shaping the future of telecoms. To gain a complete picture of the industry with insights on other MWC24 themes, download the Xona Insight Note here: MWC2024: Digital Infrastructure at Crossroads – Where Investments Are Heading Next.