Technology companies that want to innovate in the European Union find the road anything but smooth. On the one hand, unclear legislation, the difficulty of getting answers from regulators, enormous bureaucratic complexities. On the other, a chronic lack of competitiveness and funds that makes the European economy struggle compared to the United States and China. To unblock the situation, a review of the regulatory system would be needed…
This is the diagnosis that Markus Reinisch, Vice President Public Policy for Europe at Meta, offered to Adnkronos on the sidelines of the Comolake 2024 conference in Cernobbio. The Austrian manager spoke about the reasons that hinder European competitiveness and, among other things, prevent the company from launching its new multimodal artificial intelligence model Llama 3.2 in the EU. “It’s not a question of choice,” Reinisch underlines, recalling that other large companies (like Apple) also struggle to release the most innovative products.
Models like the one offered by Meta can be part of the solution, added the manager: being open source they can be freely adopted for a variety of uses, including more vertical ones. But first it is necessary to understand what has not worked so far in the EU, a theme at the center of Mario Draghi and Enrico Letta’s relations to which the President of the European Commission Ursula von der Leyen is paying great attention.
The regulatory side
Let’s start from the beginning: regulation. On the practical side, a digital reality that wants to operate in the EU must take into account at least 270 separate regulators in the various EU member states and around 100 local laws in the digital field, explains the Meta manager. “We as large companies can afford to face the difficulties and costs of compliance with various regulations. But if you are a small European company, this is impossible.”
Furthermore, it is not clear what kind of data can be used to train artificial intelligence models in the EU, nor to what extent it can be used, explains Reinisch. And then there’s the issue of timing, which is crucial given the speed at which the tech sector moves. For example – he continues – the conversations on AI between Meta and the European institutions began some time ago and will continue at least until the first quarter of 2025. The equivalent process in the United Kingdom instead took only a few weeks.
The question would be almost philosophical if it were not for its serious implications in the real economy. With the AI Act, Brussels has placed itself at the forefront of the discipline and aims to implement well-defined guardrails to reduce the risk of new applications of artificial intelligence systems. From Reinisch’s perspective, however, using only the lens of risk to evaluate these new technologies could be detrimental.
“Think about Covid: there was a need to find a balance between people’s privacy and security. What applies to health and safety also applies to productivity and innovation,” he explains, underlining the urgency of creating solutions capable of evaluating the effect of regulation on businesses. Also because the EU, which is already struggling on digital, cannot afford any further delay.
The challenges of competitiveness
For Reinisch, the countries of the Union are resigning themselves to US superiority in the development of new technologies. In Europe, the private sector invests 50 times less in AI than the United States. “I tell you that this gap will increase and we will lose even more ground,” he warns. Also because the competition doesn’t just come from the west. “We will probably also lose against China, which despite being held back by chip sanctions still manages to move faster.”
According to Reinisch, India is now competing with the EU for third place globally. “As Europeans we have missed many opportunities to create a single market, both at an industrial and regulatory level.” Not too long ago, European GDP was higher than that of the United States, he recalls, thinking back to the dawn of mobile connection: it was the Finns of Nokia, the Swedes of Ericsson and the large European telco operators who introduced 3G, while the United States chased . “In the space of 15 years we have lost first place. And as a European I say this with enormous regret.”
There is a cultural as well as economic risk in choosing not to take part in one of the greatest technological revolutions, continues the Austrian manager. “If technology is not built with thought, with the specificities of different cultures – which are represented by data as well as talents – technology will not reflect our culture,” he warns.
The solution according to Meta
So how to recover? “The inconvenient truth is that developing foundation models is incredibly expensive,” explains Reinisch, speaking of the large, versatile models that underpin modern chatbots. Meta currently invests around $40 billion a year in AI alone. “We are not even the largest investors, the value of the investments of the top four operators is around 100 billion in 2024”.
This is not a good sign for Europe, because companies and institutions would have difficulty even raising similar sums of money. Not to mention the cost and amount of energy needed to train foundation models, Reinisch continues. Therefore, given that development from scratch is “essentially impossible” for European companies, the executive reasons, European companies could exploit open source systems to skip the development phase and do “what they do best, which is to build vertical applications ”.
His reasoning expresses the company’s vision: by choosing to develop a family of open source models – therefore freely downloadable and exploitable by anyone – Meta has made a unique choice compared to those of other large tech companies in the AI field. But according to Reinisch the strategy is already paying off because the Llama models “will most likely become the standard for this industry”.
So far the model has been downloaded around 500 million times globally, he says, and among the users there are several European excellences who have used it for their own specific applications. For Europeans, “open source is a great opportunity”: the availability of the model, combined with complete control over its operation and data, are an ideal solution for European companies that want to exploit AI to become more competitive.
Meta itself has a collaboration with the Italian-French conglomerate EssilorLuxottica, together with which it developed the Ray-Ban Meta smart glasses with cameras, sensors and integrated AI system – not available in Europe precisely for reasons of regulatory uncertainty, it explains the company. Reinisch indicated them as a concrete example of the path he imagined for European companies. It is a product built with “European skills and quality, on a technological platform created in the USA”. And they are “the most successful series of AI devices” on the market, he says.