Blogs

AI at MWC 2024

This year’s Mobile World Congress (MWC) hosted by GSMA welcomed over 101 thousand attendees to the world’s largest and most influential connectivity event in Barcelona. A standout theme at this year's conference was "Humanising AI," which offered a platform to explore the practical applications of Generative AI beyond the hype.

Generative AI (Gen AI) is poised to accelerate network transformation, enhance human-machine interactions, and deliver personalized experiences for customers. The telecommunications industry, with its wealth of diverse data, is well-positioned to lead in this area. CEOs anticipate that Gen AI will significantly alter how their companies create value in the next three years, signaling a disruptive and exciting phase for the traditionally slow-paced telecom sector.

Here are 10 key takeaways from "Humanising AI" at MWC 2024 for individuals, businesses, and institutions:

  1. AI will be native to next-gen devices
    Qualcomm's Snapdragon 8 Gen 3 and Snapdragon X Elite signify a shift towards on-device AI. This advancement promises improved device performance and connectivity, opening doors to innovative applications in the healthcare and manufacturing industry which rely on real-time data processing. Notable device launches included Samsung's Galaxy Ring, Honor's Magic 6 Pro with an eye-tracking AI function linked to device control, and Deutsche Telekom's AI voice recognition smartphone. Additionally, Motorola exhibited their personalized AI assistant, Xiaomi boasted its new video-editing tools and Humane showed off its wearable, screenless AI pin where users can use the built-in laser to interact with Wi-Fi settings and media playback controls — highlighting the growing integration of AI on consumer devices.
  2. Asian telcos are pulling ahead in the AI race
    Operators in Korea and Japan are leading the rest of Asia in the latest foray into AI with the likes of SK Telecom, KT and KDDI investing heavily in capturing new opportunities with this technology. SK Telecom, which announced their “AI Pyramid” strategy already devotes approximately 12% of investments towards AI and is expected to triple to 33% by 2028. KT has also been developing AI applications and services in the robotics, logistics and AI contact centres, which cumulatively have generated over US$ 600 mn for the company, with additional plans to invest US5.4 bn by 2027 in further setting the benchmark for telcos AI competitiveness.  
  3. Telcos need to balance the role they play in AI: providers vs consumers
    There is an opportunity for the telecom industry to play a vital role in unlocking the efficiency gains from AI and for enterprise customers to harness its benefits. Be that through providing high-speed, low latency connections offered by 5G technology, or investing in and developing edge computing and ‘network-as-a-service’, telecoms companies are instrumental in enabling the safe, efficient, and effective use of AI across the globe. Being a consumer of AI, telcos can look at three levels of Gen AI tool deployment for internal benefits – general-purpose, department-specific, and task-specific tools. For those looking for a prosumer role, an evolution trend is common across three stages: deploying Gen AI Copilot for productivity increase, transforming the business by implementing AI at scale (i.e., find key business domains and group high-value use cases) and exploring AI revenue generation channels (i.e., deploying custom AI models on private or hybrid networks for enterprise customers with the right SLAs and mix of LLMs).
  4. AI deployment will need to be distributed
    With infrastructure investments advancing edge inferences with higher speeds and lower latency, there will be a unique type of convergence of 5G and AI at the edge. This demands new applications designed for the acceleration of edge content caching and edge applications with AI-native capabilities. Telcos must choose the right AI deployment model based on use cases, whether it's in a hybrid cloud, at the edge, or on devices. Telenor is taking the lead with its recent announcement of a partnership to become sovereign AI cloud partner of NVIDIA with the likes of Indosat following closely, becoming Indonesia’s first NVIDIA Cloud Provider Partner. Down the road, a key focus will be optimizing asset usage, as AI and Radio Access Network (RAN) share infrastructure. Telcos will aim to push AI inference out onto the edge to reduce costs, avoiding central cloud data centers. However, to persuade operators to invest in AI hardware for the RAN, suppliers must demonstrate that the revenue potential outweighs the additional costs of electricity and hardware wear and tear.
  5. Low hanging AI fruits for telcos will be in customer experience
    As highlighted in numerous panel sessions, telcos will see most value from AI in enhancing customer experience. Key areas include smart call routing, agent assistance, root cause analysis (RCA), next-best-action (NBA) recommendations, and agent copilots. These AI applications increase consumer stickiness — by overcoming language barriers and improve resolution capabilities — encouraging customers to stay longer in the channel, not just seeking the fastest way to talk to a live agent. Hyper-personalization is another crucial aspect, tailoring every customer journey with pure automation and augmenting the human frontline with tools that enhance agent effectiveness. Given the linguistic nature of many of these tasks, Gen AI is well-suited to address these challenges in CX.
  6. Telcos will significantly benefit from specialized LLMs
    The true value of Gen AI in many enterprises lies in the ability of foundation models to be fine-tuned for specific applications. A domain specific LLM, like the one being developed by SK Telecom and Deutsche Telekom or one focused on local language co-built by Indosat and Tech Mahindra, will bring substantial benefits. Consider the word "cell," which has different meanings in the telco and healthcare industries. In healthcare, it refers to the basic unit of life, while in telecommunications, it denotes a geographic area covered by a single base station in a cellular network. A general-purpose LLM that fails to recognize this distinction could produce highly inaccurate and contextually irrelevant outputs. In contrast, a domain specific LLM tailored for the telco industry would be able to understand and apply the correct meaning, leading to more accurate and relevant results.
  7. Open-source data will be key to creating the right LLM
    Echoing the conversation in Davos 2024, open-source data is crucial for accelerating the widespread application of AI in the industry. Operators in various regulations will need to share data openly to create effective industry specific LLMs. While it's essential for these operators to remain competitive in the data race, hoarding data is counterproductive. While CEOs and CIOs evaluate the current state of data models and architectures across their operations, they should collaborate openly to find ways to contribute data without breaching regulations, compromising privacy and security standards, or forfeiting competitive advantages. The development of an AI-ready standardized data model is an ongoing initiative at TMForum, aiming to facilitate this collaborative approach to data sharing.
  8. Telcos are increasingly looking to use SLM or LAM, not just LLMs.
    Telcos might need to look beyond LLMs to deploy AI applications effectively. For instance, if you ask ChatGPT about a telco product, it might not have the correct answers due to its lack of access to proprietary documents, leading to potential inaccuracies and poor experience. To address this, telcos are exploring the potential of smaller language models (SLMs) that, while still large (possibly several billion parameters), can be trained on focused datasets from specific domains and offer better accuracy and operational cost efficiency. Telcos might even consider adopting large action models (LAMs) to scale their generative AI capabilities which are designed to perform actions (e.g., navigating in and out of an app to complete tasks) based on their understanding, adding a layer of practical functionality to Gen AI models.  
  9. Telcos need built-in guardrails for AI use
    Telcos need to establish a comprehensive governance model for the use of AI at scale. This involves ensuring the right kind of data is brought in for AI models and scrutiny on how the models are trained. All deployments should only occur when legal departments are comfortable with the compliance and legal implications. It's crucial that the content and data used are open source and appropriately licensed, as seen with Adobe Firefly's licensing agreement with Getty Images. This ensures that the output is both safe and legal. Building ownership around the output of AI is also important. This can be achieved by adding a human element to the AI's output, ensuring that there is accountability and a sense of responsibility for the results.  
  10. AI will be increasingly native to telecom
    During the conference, major tech companies including Amazon Web Services, Nvidia, Microsoft, Nokia, and T-Mobile announced the formation of a new AI-RAN alliance. The collective goal is to integrate AI into cellular technology to further advance network performance, energy efficiency, spectrum sharing and security of RAN and mobile networks. There are three main areas of focus:
    • AI for RAN: Advancing RAN capabilities through AI to improve spectral efficiency.  
    • AI with RAN: Integrating AI and RAN to use infrastructure more efficiently and generate new AI-driven revenue streams.
    • AI on RAN: Deploying AI services at the network edge through RAN to increase operational efficiency and offer new services to mobile users.  

Although Twimbit is not a participant in MWC 2024, we have compiled these insights from interview soundbites, discussions with participating clients and publications by third-party partners. We are supporters of GSMA and their ability to drive innovative agendas to push the boundaries of telecommunications and aspire to contribute to these discussions in the future, with the aim of building a better world and delivering exceptional experiences to all stakeholders.