Germany’s Seedbox.AI is, in many ways, a company made to win the hearts of EU lawmakers: Relatively young, doing business with established players and, most importantly, developing artificial intelligence.

Co-founder Kai Kölsch tells Euractiv the company has been in the AI business since before US giant OpenAI launched its viral AI chatbot ChatGPT in 2022 – back when the technology was still simply called ‘machine learning’.

Now, like many other players in the AI field, Seedbox re-trains large language models (LLMs) developed by other companies – such as Google’s Gemini or Meta’s Llama – for its specific uses, for example a chatbot where patients can ask questions about their medication or an AI assistant for real estate appraisals.

This is where 38-year-old Kölsch sees Europe’s big chance: piggybacking on LLMs, rather than trying to clone them. “We want to drive the [AI] field forward, take part in developing technology that makes technology more efficient,” he says.

The Stuttgart native, still living in Germany’s automotive capital, likens the situation to a car that’s already on the road. Europe’s tech field shouldn’t be trying to reinvent the wheel, he suggests, but rather should focus its energy on improving the quality and efficiency of the drive.

Of course, there are still bumps in the road for achieving this vision.

Good projects for more hardware

The first big problem for Europe’s AI companies is that they do not have the same access to computing infrastructure as the US giants. Kölsch also says it would be good to have access to more of the coveted specialised chips needed for AI work – either for training or for running models.

For Seedbox, there’s a potential solution: The EU is currently building one of its AI factories in Stuttgart, designed to link startups and established industry players with the hardware to integrate AI into their offerings.

The company is already cooperating with the computing centre where the factory is being set up, which allows it to train an AI model in all 24 official EU languages. But the new specialised chips will extend the possibilities, says Kölsch.

And while he welcomes plans for far larger gigafactories also planned in the EU, Kölsch argues that these should be concentrated as tightly as possible to be useful for training new foundational models, in the vein of Google’s Gemini or Meta’s Llama.

The Commission already split up the original “AI CERN” idea into five separate gigafactories, which governments and companies are now debating to divide up even further.

Dial down the rules

Kölsch also has his doubts about EU regulation, specifically the AI Act.

He’d like to see it delayed because of the effect he reckons it’s having on established companies. “They would rather do nothing than do something wrong,” he says. “Doing nothing is the worst thing you can do right now.”

The past months have seen several calls to delay European AI rules, with a parallel discussion about whether, and how much, they should be watered down as part of the Commission’s ongoing digital simplification drive.

While the EU executive was late in delivering key supporting documents for the AI Act – and some detail is still missing – most countries are also delayed in announcing which authorities will be responsible for implementing the Act in their territory, amplifying the legal uncertainty.

“That’s where we need really clear signals”, Kölsch says. “Like, you won’t be flogged and thrown into jail if you mess up. You just have to start now.”

The bloc’s data protection law, also often singled out by critics for blocking European companies, is also set to be – at least in part – simplified.

Money, money, money

But Kölsch highlights one particular issue as central to Europe’s problems: “At the end of the day, the key point is capital,” he tells Euractiv.

There is a vast gulf between the sums of money pumped into tech in Europe vs in the US, even though the two economies are broadly comparable in size.

Key to this is venture capital for young and innovative tech companies. Such investments are higher risk but also have enormous growth potential. According to numbers from the European Investment Bank, US companies receive six to eight times more venture capital investments than European startups.

There are many, often-reported, reasons for the lack of cash. To pick just two: There are rules preventing big institutional investors like pension funds from putting money into venture because of the inherent risks. What’s more, the bloc is not a single financial sector, it’s 27 small ones, which don’t work well enough in funnelling cash to promising start-ups.

“It’s absurd that we want to finance Europe and at the same time European pension funds are investing in US bonds”, says Andreas Schwarzenbrunner from Speedinvest, an EU venture fund.

“The money is there, Europe is still very rich”, he continues – “it’s simply invested in the wrong channels.”

Schwarzenbrunner is quick to point out that Europe has made significant progress after realising it missed the boat for thirty years. There is now a working ecosystem for financing young companies. Still, European tech companies face big problems in accessing financing – both for building prototypes and then, once they have proven the tech, for expanding, optimally across the EU.

Even if they manage to do so, US Big Tech is lurking with many billions in cash which they will gladly use to snap up (and close down) promising start-ups.

Kölsch says that Seedbox.AI is already being courted by non-European companies like AMD and Nvidia while European ones are failing to keep up with the speed of technical innovation.

“No company in Germany… really understands what we do”, he says. “Who understands us is the Americans.”

Talent follows money

This connects to another problem that EU and national governments have wrestled with for years: talent.

Seedbox.AI currently has 15 employees and is recruiting a new AI engineer – as are other European companies, with many struggling to find staff.

To boost the potential workforce, the EU made 2023 a ‘Year of skills’, with a strategy on a “Union of Skills” that aimed for one in three university students to enrol in STEM degrees by 2030 – which are foundational for AI development.

“By offering world-class education and research infrastructure, competitive career prospects, and a supportive regulatory and funding environment, the EU can become a destination of choice for the brightest minds,” the skills strategy reads.

But it’s hard to compete when Big Tech is dangling eye-watering salaries to reel in AI talent. Top researchers have reportedly been offered first-year pay packages of more than €100 million to join Meta.

Meanwhile, the word “salary” does not appear in the EU’s Union of Skills strategy.

“We can’t allow ourselves to complain that we can’t educate talent,” says Kölsch. “Talent just goes to where the capital is.”

(nl, jp, ow)