ChatGPT, Gemini, DeepSeek, Le Chat... These are all now familiar names, synonymous with a technology that, in just a few months, has taken its place at the heart of public, economic and political debate. Since the launch of ChatGPT at the end of 2022, artificial intelligence has gone a step further, democratising access to tools that were once reserved for a handful of insiders.

For many, this advance constitutes a revolution. Luc Julia, one of the designers of the voice assistant Siri and now chief scientific officer at Renault Group, prefers to talk about evolution: "Generative AI is simply the logical extension of machine learning and deep learning. What's changing today is the trillions of data available--that's almost the whole internet--and above all the use: anyone can use it without training."

Far from Hollywood fantasies, AI remains, he insists, "a toolbox", useful, certainly, but only for certain specific tasks.

An unavoidable economic reality

In Luxembourg, lucidity prevails, without giving in to scepticism. "AI is no longer a myth, but a reality," says economy minister (DP), who aims to make it a lever for competitiveness while taking care not to leave anyone by the wayside. "A business that is not digitalised could no longer exist today. The same will be true tomorrow for those that ignore AI", he asserts, thus justifying the recent of aid for the digitalisation of SMEs.

As the movement has begun. According to the latest economic barometer from the Chamber of Commerce, 63% of Luxembourg businesses believe that their model will be impacted by emerging technologies and plan to invest in AI within three years, points out its director general, .

Economist Philippe Aghion even sees it as a transformation of historic proportions: "AI is not just automating the production of goods and services. It is also accelerating the production of ideas. Its potential impact is comparable to that of electricity or computing."

The chairman of the Chamber of Commerce's AI working group and head of Proximus Luxembourg, puts a figure on this opportunity: a gain of nine percentage points of GDP over the next five years, provided there is massive investment, particularly in the financial and public sectors. Conversely, a lack of ambition could prove costly.

A targeted strategy

Far from wanting to compete with the giants of the sector, Luxembourg is banking on a targeted strategy. "We need to build smaller, specialised models that are adapted to our size and our needs," says Jacques Thill, government adviser at the ministry of the economy. A position shared by (ABBL), who stresses the importance of "pragmatic digital sovereignty", particularly when it comes to data governance.

The country's expertise in digital infrastructure, with programmes such as by Luxinnovation or upcoming projects such as as well as the , are assets. But the key also lies in human resources. The "talent drain" is a major brake, warns Gérard Hoffman. Luc Julia confirms this: "All the AI managers in the big American companies are French."

Another avenue mentioned was that of regulatory experimentation environments (or ), defended in particular by Gérard Hoffman and Carlo Thelen. "By offering a favourable and pragmatic environment, Luxembourg can become the gateway for companies from third countries wishing to test the compliance of their AI products", the latter argues.

Europe faces the challenge of coordination

At the continental level, the observation is identical. For the president of the European Investment Bank, , "AI is an incredible economic opportunity". But the European Union will only be able to make an impact if it moves forward "as a team", she warns. "Europe has a card to play, provided it invests massively, simplifies its rules and strengthens the integration of its market", argues the former Spanish minister, drawing on the recommendations of the .

Gérard Hoffman, for his part, argued for a focus on vertical markets where Europe already has expertise, such as health tech or technologies for the public sector.

We still need to close the investment gap that separates Europe from the United States and China. In this regard, Thelen cites the US Stargate programme, with $500bn earmarked for AI-related infrastructure, far ahead of the €200bn announced by the EU.

Regulating without putting the brakes on

But it is undoubtedly on the regulatory front that the debate is most heated. The European IA Act is raising many questions about the balance to be struck between security and innovation. Jacques Thill acknowledges that being the first continent to legislate in this area exposes Europe to criticism.

In addition to infrastructure, training and funding, the issue of governance is crucial. Who decides how AI is used? Who controls biases, hallucinations and systemic risks? If the European model wants to distance itself from the permissive approach of the United States or the authoritarian control of China, it must build a third way, based on transparency, responsibility and trust.

"This is good regulation", notes Ananda Kautz. "But it will need to ensure that compliance costs do not hamper AI adoption, particularly in an already heavily regulated financial sector."

Another source of concern for Philippe Aghion: the risk of monopoly. During the IT revolution, he recalls, "super-firms" initially stimulated growth, before holding back the entry of new companies by becoming monopolies. He fears a similar scenario with AI. "The cloud is already dominated by three giants - Amazon, Google and Microsoft - while Nvidia reigns alone in GPUs", he analyses. "We need to get away from the 'winner takes all' logic and open up the market to a diversity of players." This presupposes proactive regulation, but also a strengthening of the European institutions capable of enforcing it.

Promises to be tempered

But the promises of AI should not blind us to its limitations. Luc Julia points out that there is still room for improvement in the accuracy of generative AI: a study by the University of Hong Kong, published in 2023, puts it at 64%. And the rest? "Hallucinations", like the case law invented by an AI in a legal memo, used by a lawyer who was dismissed for this reason, he recounts.

Another major issue: environmental impact. "A 100-word e-mail generated by AI consumes as much as a bottle of water," he warns. "While Luxembourg is fortunate to have sufficient resources, this is not the case everywhere in the world."

As the race for artificial intelligence intensifies, Economy Day has opened up a debate that is as necessary as it is complex. While Luxembourg, like Europe, intends to play its card, the equation remains delicate: encouraging innovation without giving in to technological intoxication, regulating without hindering, investing without excluding. It's a line of conduct that must be followed without wavering.

This article was originally published in .