Europe’s urgent drive to tackle AI sovereignty challenge

Europe’s generative AI startup scene is slowly taking shape, although it hasn’t reached the level of fame enjoyed by North American counterparts like OpenAI, Anthropic, or Cohere. Just recently, Germany’s Aleph Alpha made headlines by securing a whopping €460 million in funding, marking one of the biggest funding rounds ever for a European AI company.

The European tech realm greeted the investment news with a touch of excitement. Despite the spotlight on how the EU and the UK navigate tech regulations, there’s been a relative lack of emphasis on the bloc’s strategies to bolster innovation in artificial intelligence and mitigate the potential of trailing behind in the ever-evolving tech landscape. At a press conference discussing the investment, Germany’s Vice Chancellor and Minister for Economic Affairs, Robert Habeck, emphasized the need to back local AI businesses.

“The thought of having our own sovereignty in the AI sector is extremely important. If Europe has the best regulation but no European companies, we haven’t won much,” Habeck said.

Also Read: ESA’s big plan to eliminate new space debris within a decade

The importance of transparency, traceability, and sovereignty

During that press conference, Jonas Andrulis, the founder and CEO of Aleph Alpha, mentioned that the investors involved in the recent funding round, including familiar names like SAP, Bosch Ventures, and the owners of Lidl, were all longstanding partners of the company. Interestingly, nearly all the funding, with only a minor contribution from Hewlett Packard, came from European investors or grants.

Andrulis stressed the significance of transparency, traceability, and sovereignty from the get-go in their research. This underscores ethical considerations that could distinguish a European LLM, along with meeting geopolitical goals.

Aleph Alpha is working on a big language model (LLM) akin to OpenAI’s GPT-4, but with a focus on catering to corporations and governments rather than individual consumers. However, there are some notable differences between the two companies — OpenAI boasts a workforce of 1,200 employees, while Aleph Alpha has a team of 61. Additionally, the former has successfully raised over €11 billion in funding.

Yet, the game might change for Aleph Alpha with the establishment of the €2 billion Innovation Park Artificial Intelligence (Ipai) in their hometown of Heilbronn in southwest Germany. This development could provide the startup with the boost it requires to compete on a more even footing.

AI startups raking in big bucks

Ipai’s construction is set to wrap up in 2027, with the capacity to house 5,000 people. The Dieter Schwarz Foundation, which played a part in Aleph Alpha’s recent funding round, is backing this project. Established in 2019, Aleph Alpha isn’t a rookie in the field. In 2021, predating the investment frenzy sparked by ChatGPT, the company secured €23 million in a funding round spearheaded by Lakestar Advisors.

Of course, that sum has been dwarfed by billion-dollar figures across the Atlantic since then. Nevertheless, Aleph Alpha did manage to pull in an additional €100 million, supported by Nvidia and others, just this June. And it’s not just the German startup scoring big bucks. Just a few weeks after its founding in May this year, France’s Mistral AI pulled in a whopping €133 million in what’s said to be the largest-ever seed round for a European startup.

Mistral, the French AI developer, claims that its Large Language Model excels in low latency, text summarization, classification, text completion, and code completion. The startup has chosen to unveil Mistral 7B with the Apache 2.0 license, allowing unrestricted use and reproduction as long as proper attribution is provided.

They also mentioned their belief that the future will witness a multitude of specialized models, each tailored for specific tasks, compressed to the maximum extent, and linked to specific modalities. Mistral, based in Paris, grabbed attention in June by becoming the purportedly largest-ever seeded startup in Europe, securing €105 million in a round spearheaded by Lightspeed Venture Partners.

Despite the seemingly Herculean strides made in the last three months, it’s worth noting that all three founders of the company hail from Google’s DeepMind or Meta. The Mistral 7B, with its 7.3 billion parameters, is said to outshine larger models like Meta’s 13 billion parameter Lama 2 while demanding less computing power.

Giving Aleph Alpha a run for its money in the race for the European GenAI crown, Mistral AI is also crafting an LLM tailored for enterprises. Interestingly, their inaugural model, Mistral 7B, comes with a completely free-to-use tag. When making pitches to investors, Mistral, founded by ex-Google and Meta employees, apparently underscored the significance, labeling it a “major geopolitical issue,” that Europe lacked a formidable contender in the field of generative AI.

Also Read: ChatGPT-4 Turbo explained: What makes it the biggest update since launch

Netherlands aiming for a transparent, fair, and verifiable alternative

At the same time, it’s not just the German government making moves to strengthen local generative AI capabilities. The Netherlands has also kicked off the creation of its own in-house LLM, aiming to offer a “transparent, fair, and verifiable” alternative in the realm of GenAI.

The upcoming Dutch Large Language Model, named GPT-NL, is set to be an open model. Its creators emphasize that it will be accessible for everyone to examine the inner workings of the software and understand how the AI arrives at specific conclusions. The development of this AI involves collaboration between the research organization TNO, the Netherlands Forensic Institute, and the IT cooperative SURF.

In the coming year, the collaborators will concentrate on building and training the Large Language Model (LLM). Once this phase is complete, it will be open for use and testing. GPT-NL will be linked to the national supercomputer Snellius, supplying the necessary processing power to make the model operational.

Quite fittingly, the introduction of GPT-NL happened to align with the world’s first AI Safety Summit. This star-studded event, held at Bletchley Park in the UK, delved into strategies to address AI risks through globally coordinated efforts.

Only a few days before the summit, UK Prime Minister Rishi Sunak revealed the rollout of an AI chatbot designed to assist the public in handling tax payments and accessing pensions. However, in contrast to GPT-NL, the technology powering this experimental service is overseen by OpenAI — symbolizing the UK’s more hands-off approach to managing big tech compared to the EU.