Germany's Seedbox.ai is a company created in many ways to win the hearts of EU lawmakers.
Co-founder Kai Kölsch told Euractiv that the company is involved in the AI business. This is when technology was still called “machine learning” even before Giant Openai launched the viral AI chatbot ChatGPT in 2022.
Now, like many other players in the AI field, Seedbox retrains large-scale language models (LLMs) developed by other companies such as Google's Gemini and Meta's Llama. For example, a chatbot where patients can ask questions about AI assistants for drugs or real estate assessments.
This is where 38-year-old Kersch sees a huge opportunity in Europe. It's piggyback in LLMS rather than trying to clone them. “We want to drive [AI] Please join in developing technology that makes technology more efficient,” he says.
The Stuttgart native, who still lives in the German car capital, compares the situation to a vehicle already on the road. He suggests that the European high-tech sector should not try to reinvent the wheels, rather that its energy should be focused on improving the quality and efficiency of drives.
Of course, there are still bumps on the roads to achieve this vision.
Good project for more hardware
The first major problem for European AI companies is that they have no access to the same computing infrastructure as the US giants. Kölsch also says that it's good to have access to the coveted specialized tips needed for AI work, either in training or running models.
For Seedbox, there is a potential solution. The EU is currently building one of its AI factories in Stuttgart, designing it to link startups and established industry players with hardware to deliver AI.
The company is working with the computing centre where the factory is already installed, allowing it to train AI models in all 24 official EU languages. But new specialized tips expand possibilities, says Kelsch.
And while he welcomes plans for a much larger Gigafactory planned in the EU, Kersch argues that in the vein of Google's Gemini or Meta's llama, we should focus as much as possible to help train new underlying models.
The committee has already split the original “AI CERN” idea into five separate Gigafacties, and the government and businesses are currently discussing it to split further.
Dial down the rules
Kersch also questions EU regulations, particularly AI law.
He hopes it is behind because of the effects he thinks he is giving to established companies. “They would rather do nothing but do something wrong,” he says. “Doing nothing is the worst thing you can do right now.”
In the past few months, there have been several calls to delay European AI rules, and parallel debates as to whether they should be watered down as part of the committee's ongoing digital simplification drive.
EU officials have been late in filing key supporting documents for AI law, but some details are still missing, but most countries are also late to amplify legal uncertainty by announcing which authorities are responsible for enforcing the law on their territory..
“That's where you really need a clear signal,” says Kelsch. “If you ruin something, for example, you won't get whipped and thrown into prison. You have to start now.”
The data protection laws for blocks selected by critics to block European companies will also be at least partially simplified.
Money, money, money
However, Kersch emphasizes certain issues as the centre of European issues. “The key point, at the end of the day, is capital,” he tells Euractiv.
Despite the size of the two economies being roughly equal, there are money pumped into technology in the US and vast bays in the US.
The key to this is venture capital for young, innovative high-tech companies. Such investments are risky, but they also have great growth potential. According to figures from the European Investment Bank, US companies receive six to eight times more venture capital investment than European startups.
There are many, often reported reasons for cash shortages. To choose just two, there are rules that prevent large institutional investors, such as pension funds, from putting money into ventures for inherent risks. Furthermore, the bloc is not a single financial sector, but a small sector of 27, which does not function well enough to concentrate cash on promising startups.
“It's ridiculous that it's funding Europe and at the same time European pension funds are investing in US bonds,” said Andreas Schwarzenbrunner of SpeedInvest, an EU venture fund.
“The money is there and Europe is still very rich,” he continues – “it's simply invested in the wrong channel.”
Schwarzenbrunner quickly points out that Europe has made great strides after realizing he missed the boat for 30 years. Now there is a practical ecosystem for funding young businesses. Still, European tech companies face major challenges in accessing funding, for building prototypes and to expand optimally across the EU after they prove their technology.
Even if they manage to do so, big US technology is hiding in billions of cash.
Kölsch says Seedbox.ai is already courted by non-European companies like AMD and Nvidia, but European companies have not kept up to the speed of innovation.
“There's no company in Germany… I really understand what we're doing,” he says. “The ones who understand us are Americans.”
Talent follows money
This leads to another problem that the EU and the central government have struggled over the years: talent.
Seedbox.ai currently has 15 employees and is looking for new AI engineers, just like other European companies. Many people struggle to find staff.
To boost the potential workforce, the EU made 2023 a “year of skills” with its strategy on “coalition of skills” aimed at enrolling in a STEM degree by 2030. This became the basis for AI development.
“By providing world-class education and research infrastructure, competitive career outlook, and a supportive regulatory and financial environment, the EU could be the best destination for the brightest mind,” reads the Skills Strategy.
But when Big Tech gets caught up in AI talent and hangs his pay to look away, it's hard to compete. Top researchers reportedly are being offered a one-year salary package of over 100 million euros to participate in the meta.
On the other hand, the term “salary” does not appear in the EU's Skill Strategy Coalition.
“You can't complain about not being able to educate talent,” says Kelsch. “Talents just go to places in the capital.”
(NL, JP, OW)
