AI: A true game changer for connectivity networks

Over the last two years, we have seen the launch of a new industrial revolution with an ongoing exponential use of AI and Generative AI. AI has already started to spread across all the various sectors of activities and into our private lives. The AI revolution creates huge opportunities for the telecom sector to become more efficient, more competitive with smarter networks, and more innovative with customised offers combining AI and 5G slicing. This also creates huge responsibilities to ensure AI development and usage are responsible, inclusive, and compliant with our green commitments.

As far as connectivity networks are concerned, the AI effect is twofold: it will substantially transform the way we deploy and run our networks with greater efficiency and customer satisfaction, and at the same time it will require additional investments in innovation to ensure our networks are fit for the increased use of AI in our society.

AI for networks: towards smarter networks for the benefit of all

AI is going to be a game changer for networks. They will become increasingly smarter and customised to various demands. Thanks to AI, they will run more efficiently, be more agile and flexible, with enhanced cost optimisation.

And this is not theory but already very concrete in practice! We have at Orange many use cases that are already deployed or under test in our networks. This includes for instance to use AI for

• Enhancing customer experience; AI can significantly improve customer experience by providing personalised services and instant support, thanks to AI-driven chatbots, enhancing user satisfaction and reducing response times, or thanks to personalised recommendations, enabling tailored service offerings that meet individual needs.
• Optimising network investment choices, by designating the best network (antennas) location, and/or improved network capacity planning to optimise traffic flow.
• Network management, improving field operations, performing predictive maintenance, or summarising trouble tickets, helping technicians to fix issues.
• Fighting against cyber-attacks – which are themselves increased with AI. AI can be used to enhance cybersecurity protocols, enabling proactive threat detection and response, for instance with an AI tool in a anti DDoS system, or to protect submarine cables by monitoring with an AI sound tracker all types of activities ongoing around cables.
• Greener connectivity networks. While AI is raising huge challenges in terms of energy consumption (notably linked to data centres’ energy and water consumption) that will have to be tackled, it can also enable the reduction in emissions. For instance, by using AI systems to monitor the energy consumption of our routers on international links, we have already managed to reduce this consumption by 12%.

To summarise, using AI in our networks translates into a better customer experience, especially in terms of service reliability, reduced downtime, optimised performance, and more efficient customer support. Our business customers also benefit from more robust and secure networks tailored to their specific needs, including their own AI tools.

To reap the most benefits of AI powered connectivity networks, and in addition to the cyber and green aforementioned challenges, we also need to address the following issues:

• Improve EU citizens and employees’ skills on AI, which requires specific and adapted training. In Orange a wide range of trainings is available for our employees as well as a specific service named Dinotoo[1], which is a secure generative AI tool based on the most powerful LLMs with the utmost security and data protection available for all. Since its launch in January 2024, over 56,000 distinct employees used it. Based on its experience, Orange Business launched recently Live Intelligence: a range of plug-and-play GenAI solutions for businesses.
• Risk of hallucinations, which calls for testing AI systems robustly before any wide implementation.
• Data management and usage; entities need to adapt themselves to be able to manage and use the data at their disposal in the company in the most efficient way and to maintain high data quality. This is one of the most difficult tasks to achieve; “Data Democracy” is our dedicated multi-year programme in that domain.

On the other hand, the AI revolution cannot take place without connectivity and robust networks able to convey AI-driven traffic.

 Networks for AI: new challenges to cope with specific AI-driven traffic

AI will require more innovation in networks: we expect an increase in the volume and strong changes in the type of traffic we convey; according to OMDIA by 2030, two thirds of traffic will support AI based workloads. With AI running on diverse equipment, from devices to clouds, there will be more diverse traffic patterns and more upstream video traffic – supporting new video traffic in both directions is rather new for our networks, and more demanding requirements in terms of quality of service (including on demand), and in particular lower latency and/or high reliability.

If we look for instance at an AI powered virtual personal assistant, we don’t speak only about conveying text but conveying multi modal interactions (text, images, videos) allowing immersive and interactive experiences thanks to low latency with real time processing and immediate responses. Specific quality (latency, real time, etc) requirements will also be required for instance in case of immersive interactive games, for AI powered smart cities, or AI-based visual quality inspection in a factory line. AI native applications will leverage a data-driven, knowledge-based ecosystem, where information is consumed instantly and dynamically generated to realise new AI-based functionalities or augment and replace static mechanisms.

Such types of usage strain networks requiring innovations in infrastructure, as well as traffic optimisation, and AI-driven resource management. With this evolution, edge computing may emerge as a potential solution to tackle some of the challenges: by processing data closer to the source, edge computing can indeed significantly reduce latency and bandwidth usage. Edge computing is also a way to strengthen security and data sovereignty, data being stored at local level. A combination of 5G network slicing and edge computing seems to be one of the possible solutions to support AI-driven innovative services.

This is why we call for a new IPCEI as a complement to the current CIS IPCEI to boost investment in edge computing solutions for AI use cases, with relevant budget for a pan-European sovereign coverage.

Finally, to facilitate global AI usage in a responsible way, which means addressing CO2 efficiency, data protection, AI safety all in a transparent and configurable way, specific AI routing mechanisms will need to be invented and deployed. Interoperability and routing among different AI systems and networks will be required at some point.

Standardisation efforts will be needed gathering industry stakeholders as well as work on open common APIs to encourage innovation and collaboration among developers and businesses.

On the latter, the sector has already started the journey thanks to the GSMA Open Gateway project, the CAMARA project[2] – an open source project within the Linux Foundation to define, develop and test the APIs, and ultimately the joint venture[3] that Orange together with some major telecom providers and vendor will launch to accelerate the adoption and innovation of network APIs.

AI is there, let’s ensure the EU collectively benefits from it in a responsible way

AI will substantially transform our society. Orange is committed to playing a pivotal role in the responsible development and deployment of AI technologies, by transforming our networks to support this evolution. This requires an adequate European framework supporting private investment and allowing a fair return on investment. It also requires ensuring a harmonised implementation of rules impacting data and AI usage between Member States so that multi-national companies like Orange do not face further difficulty in making pan-European innovation in infrastructure and systems.

More globally the EU cannot afford to miss the AI revolution; it’s the opportunity to strengthen European competitiveness, growth, and innovation. While the AI Act is now entering its implementation phase, it has become urgent to boost innovation in the EU and take-up in responsible AI. As mentioned by Mr Draghi in his report on the Future of European Competitiveness “The EU must have the ambition to be a leader in developing AI for its sectors of strength, regain and retain control over data and sensitive cloud services, and develop a robust financial and talent flywheel to support innovation in computing and AI.” We strongly hope this new European mandate will allow such ambition to materialise.

[1] https://hellofuture.orange.com/en/orange-is-developing-secure-and-streamlined-generative-ai-for-its-employees/

[2] https://camaraproject.org/

[3] https://newsroom.orange.com/global-telecom-leaders-join-forces-to-redefine-the-industry-with-network-apis/