March seems a long way off. Five years after the presentation of Luxembourg's first artificial intelligence strategy by the then prime minister, Xavier Bettel (DP), and the then economy minister, Étienne Schneider (LSAP), the update of this document is all the more eagerly awaited as other leading countries are not waiting for Luxembourg. Or have not waited for the country to move forward.

Two years before Luxembourg's first strategy, China had already decided to become the world leader in AI by 2030, capitalising on a huge amount of data thanks to its population and its well-developed digital ecosystem. Today, six of the ten largest holders of patents on generative AI are Chinese (Tencent Holdings, Ping An Insurance Group, Baidu, Chinese Academy of Sciences, Alibaba Group and Bytedance) or even thirteen if we take the 20 largest groups (BBK Electronics, Netease, Huawei, Tsinghua University, Zhejiang University, China Mobile, State Grid), and yet the world continues to talk about the Americans - and often about OpenAI's ChatGPT alone - while IBM is the American leader in patents, ahead of Alphabet, Microsoft and Adobe. And it's a safe bet that the big tech companies are going to use their cash to buy everything they lack in the value chain at some point.

Another player to keep an eye on is Singapore, which has a population of 6m and launched its Smart Initiative in 2014, followed in 2017 by AI Singapore, which was endowed with €110m to encourage research by three players (the Agency for Science, Technology and Research, National University of Singapore and Nanyang Technological University). In addition to its investments in advanced data centres and 5G connectivity, ensuring a solid foundation for AI projects, and then the creation of regulatory sandboxes to allow companies to test emerging technologies in a controlled environment, Singapore published its "AI Governance Model", one of the first of its kind in the world, from 2020.

Luxembourg's size and existing digital infrastructure are structural assets that will contribute to coordinated government-wide efforts in AI, data and quantum technologies.

Elisabeth Margue, deputy minister responsible for media and connectivity (CSV)

How can Luxembourg position itself in the face of these billions in investment? By thinking "small", as a "small country". And in this sector, that's far from a bad idea. Especially now that the country has a first-rate infrastructure (network and data centres) and that on average 50% of funding comes from the European Union.

“Luxembourg's size and existing digital infrastructure are structural assets that will contribute to coordinated government-wide efforts in AI, data and quantum technologies,” the deputy minister responsible for media and connectivity, Elisabeth Margue (CSV), stated at the second Data Summit Luxembourg in December. “Thanks to these assets, it will be possible to make progress and experiment with solutions faster than the major nations. Our aim is not simply to be first, but to deliver tangible benefits to our citizens and businesses as quickly as possible. In 2023, our shared vision was to develop three integrated strategies to prepare the country for 2030.”

Margue was already aware that Luxembourg had been selected by the EuroHPC Joint Undertaking to host a supercomputer optimised for artificial intelligence, called "MeluXina-AI", and to set up an AI factory.

MeluXina-AI will be operated by Luxprovide, which is already responsible for the MeluXina and MeluXina-Q supercomputers (the future quantum computer), and will be integrated into the EuroHPC network, according to the press release issued by the government following the weekly cabinet meeting. At the same time, Luxinnovation will coordinate the activities of the AI factory, which will use this infrastructure to stimulate the national AI ecosystem.

And therein lies the 'secret': Meluxina HPC, Meluxina-AI and Meluxina-Q are not three Ferraris lined up in front of a luxury hotel for a display of wealth, but a technological package in which each can feed the other with the lowest latency in the world - and latency, in this world of billions of millions of calculations, needs to be as small as possible.

To solve a very difficult problem, such as predicting the planet's climate in 50 years' time or designing complex new drugs, let's imagine that HPC is like a huge computing factory with thousands of workers (processors and graphics cards). This factory is capable of performing thousands of calculations in parallel very, very quickly. In this huge factory, we hire "intelligent super foremen" (AI) who can read the data, recognise patterns and learn from the past. These foremen do not replace the brute force of the factory, but they help it to work much more efficiently. Quantum computing is a bit like hiring a small team of specialists with radically new tools. These specialists (quantum computers) are not there to do everything for the factory and AI, because they are still rare, expensive and fragile. But they can provide unprecedented solutions to certain particularly complex problems that are impossible or too time-consuming to solve, even with a giant factory.

Five and a half advantages

When the three are combined, AI can quickly direct HPCs toward the most useful operations, while HPCs can in turn train and improve AI models more quickly; quantum computers can be called upon without waiting for long data transfers to solve a specific part of the problem. And because quantum computers are sensitive to 'noise' and hardware imperfections, HPCs can simulate quantum circuits and model the 'noise' and apply corrective noise measures itself. An HPC can also take an active role by recovering the results of the quantum calculation, detecting deviations linked to noise and sending an optimised sequence to correct this noise. This virtuous loop ends up making the results more reliable.

The second advantage, which also takes on particular significance in an international context where cyber-attacks are exploding, is that with the three technologies working together, it is easier to implement security measures and guarantee the confidentiality of data (which is particularly sensitive in the industrial, scientific and medical fields).

Third advantage: having these technologies on the same site creates a centre of excellence. This attracts researchers, engineers, companies and investors. The latter take advantage of the synergy to develop new products, services or algorithms. Luxembourg can thus position itself as a benchmark European and global hub, stimulating research and the local economy.

Industrial, scientific or institutional projects can move more quickly from idea to implementation. Experimentation, testing and iteration cycles are accelerated by this technological proximity. The time saved ultimately translates into financial gains, competitive advantages and a faster impact on society (for example, finding innovative medicines more quickly, improving weather forecasts, optimising complex supply chains, etc.).

Finally, although not all projects need it in the same way, Luxembourg also has a sovereign cloud offering, where data is inaccessible with Luxconnect, for the most radical offering. And a well-positioned research ecosystem, from List to SnT and the University of Luxembourg. Not to mention networking at European level.

By laying the foundations for intelligently connected, contextualised, rapidly available and secure data, Luxembourg aims to position itself in the short term as one of Europe's leaders in the field of data, said the minister for digitalisation and minister for research and higher education, Stéphanie Obertin (DP), also speaking at the second Data Summit Luxembourg at the European Convention Center.

Read the original French-language version of this report here / lire en français