The European Union has big plans to support AI startups within its borders. This includes providing access to processing power on the bloc’s supercomputers for model training, which was first announced in September and has recently entered a pilot phase with French company Mistral AI participating. However, one key lesson learned from this program is the need to provide dedicated support and training for AI startups to make the most out of the EU’s high performance computing resources.
“One of the things that we have seen is the need, not only to provide access but, to provide facility — especially skills, knowledge and experience that we have in the hosting centres — on how this access can be not only facilitated but to develop training algorithms that are using the best of the architecture and the computing power that is available right now in each supercomputing center and in our machines,” said an EU official speaking during a press briefing today.
The EU’s plan involves setting up “centers of excellence” to support AI startups in developing dedicated algorithms that can run on their supercomputers.
A majority of AI startups are more familiar with using compute hardware from US hyperscalers for training their models, rather than utilizing the processing power offered by supercomputers. To address this, the EU is augmenting their high performance computing program with dedicated support, as confirmed by EU officials during a press briefing before the official inauguration of MareNostrum 5, a pre-exascale supercomputer located in Spain.
“We’re developing facilities for our SMEs to be able to understand how to best use the supercomputers and how to access the supercomputers and how to parallelize their algorithms in the case of AI to be able to develop their models,” said a Commission official. “As of 2024, we expect much more of these kinds of approaches than we have right now.”
The EU recognizes the strategic importance of AI and aims to provide innovation opportunities for their SMEs and startups to effectively utilize their public infrastructure in developing safe, trustworthy, and ethical AI algorithms.
Another EU official confirmed that an “AI support center” is in the works, with a special track dedicated to helping SMEs and startups make the most out of the EU’s supercomputing resources. This is in light of the fact that the AI community has not heavily used supercomputers in the past decade and needs assistance in engaging with this type of technology.
“In many cases, the AI community comes with a vast knowledge on how many GPUs can be fitted into one box. However, we have a lot of boxes with GPUs on our supercomputers, and there are additional skill sets and support needed to fully utilize the supercomputers,” the official explained.
Recent years have seen the EU significantly increasing its investment in supercomputers, growing their hardware to a cluster of eight machines located throughout Europe. These machines will be interconnected via terabit networks to create a federated supercomputing resource, which will be accessible through cloud computing to users all over Europe.
The first exascale supercomputers are expected to come online in the next few years, one in Germany by next year and another in France by 2025. The EU also has plans for investing in quantum computing and aims to combine these two types of hardware – supercomputers and quantum simulators – to create a hybrid resource.
Some of the projects currently being developed on the EU’s high performance computing resources include Destination Earth, which simulates Earth’s ecosystems to better model climate change and weather systems; and a project to create a digital twin of the human body, which has the potential to support medical science through drug development and personalized medicine.
The EU is now placing a growing emphasis on using their supercomputers for AI-specific projects and has recently announced a “Large AI grand challenge.” This competition is geared towards European AI startups experienced in working with large-scale models, with the aim of selecting up to four promising startups to receive a total of four million hours of supercomputing access and a €1 million prize.
While the EU has had a program in place to provide industry users with access to core hours of supercomputing resources, they are now ramping up their efforts to support commercial AI. The goal is to leverage their supercomputing network to strengthen the capacity for developing general-purpose AI in Europe. This is evident in the early involvement of Mistral, a French AI startup competing with US giants like OpenAI and offering “open assets” for use.
Although it’s still early days for the EU’s ‘supercomputing for AI’ program, it’s clear that they are committed to nurturing a thriving AI ecosystem within their borders. Their strategic investment in high performance computing and plans to develop more advanced AI-focused supercomputers by 2024 demonstrate their dedication to closing the gap between Europe and the US in terms of AI development. With the EU’s support, it’s likely that more AI startups will have the opportunity to succeed and compete on a global scale.