How can we leverage our expertise on a European level so as to ensure European sovereignty in AI?

We’ll be able to massively increase our AI expertise in Europe and then secure European sovereignty in AI if we focus on three things: data quality, sustainability and a federated AI ecosystem for Europe.

1. Data  

The first success factor is data. Data is the lifeblood of artificial intelligence. The better the data we feed into our AI, the better that AI will be. So we need the best data we can get – in terms of both quality and volume.  

We also need heterogeneous data. If we look at the coronavirus pandemic, it’s clear that data from one country alone is far less useful in, say, forecasting, than if we could use data from a number of European countries.

Don’t forget that AI is not an end in itself. It always needs an application domain to achieve its desired effect and benefit for our society. That’s why we need a common purpose and approach to the domain-specific supply of data for our AI projects.  

We all know that AI processes are only as good as the data they’re supplied with.

2. Sustainability of AI Applications

The second key success factor is the sustainability of AI applications. When I say “sustainability,” I mean it in two different, but complementary domains: data and hardware.

First, data is a resource and we should be very mindful about how we handle it. As we all know, every search in a search engine results in CO2 emissions.

So we have to keep in mind: Not only is data very valuable, but collecting it and storing it produces CO2. The large high-performance computers for AI applications emit significant amounts of CO2 – and those levels are on the rise.

We have to keep this in mind, and that’s where we need to start. How do we do that? The most resource-efficient way possible for us to handle data is not to treat it in isolation, but to learn to take its entire life cycle into consideration.

What data am I even collecting? What’s the smartest and most efficient way to gather it?  What data should I store? And what data should I be regenerating every time if I want my process to be as sustainable as possible? What can we do to improve the sustainability, energy-efficiency and adaptability of the algorithms?

The second meaning of sustainability is that the hardware must be green and sustainable as well.

We need a sustainable infrastructure that consumes less energy and uses resources responsibly. At Fraunhofer we’re doing a lot of work in the field of Green ICT.

This is led by the Research Fab Microelectronics Germany, where Fraunhofer works very closely with the Leibniz Association. We look at all aspects of the value chain: from the technological processes, components and circuits to the system’s entire life cycle.

But even that is only the beginning. It’s vital here, too, that we strive for European collaboration. For example, Germany, France and Belgium have established a strong collaboration regarding hardware for next-generation computing.

3. A Federated Data Supply Concept

The third factor for achieving AI excellence in Europe is a bespoke, federated data supply concept. We can’t base this on the purely capitalist models of the US or on China’s state-controlled concepts.

In line with the federalist idea of Europe, we need a federated system for data in Europe that allows for optimum collaboration while retaining regional sovereignty over the data.

Through regional bodies of data – whether for a municipality, a group of companies or critical infrastructure such as an airport – we can ensure data security, rapid adaptability and simultaneously work more energy-efficiently.

We need to think European and embrace collaboration

To boost these three success factors – data quality, sustainability and a federated data concept) – we need to think European and embrace collaboration. We need investment in applied research in close collaboration with industry to pave the way for successful AI applications.

Prof. Dr. Alexander Martin, Director Fraunhofer IIS