If Germany is so technologically advanced, why are German startups so underrepresented in the 100 best AI startups in the world? According to Applied AI, in 2019 just one Berlin-based startup (Twenty Billion Neurons) secured a place on the CB Insights AI 100 list.
We wanted to get down to the facts by talking with three founders/co-founders of German startups working in the field of AI. What’s new, what’s joyful and what’s downright frustrating about being an AI-focused startup in Deutschland?
The first person joining us is Matthias Auf der Mauer from AiSight, a startup which solves the issue of machine failures, which account for 20% of the total cost of production in the industry — something which in Germany alone costs companies 44 billion euros a year.
Also here to chat with us is Tobias Martens from Whoelse.ai, a startup which offers a standardized language for conversational AIs.
Finally, we’re joined by Achim Hoth of Vialytics. Vialytics uses smartphones to help cities track road damage so they can react quickly to problems and save money by being proactive.
Achim (Vialytics): Well, I guess the EU — and within the EU, Germany especially — is very concerned about data privacy. Since data is the basis for almost all machine learning today, that is an important factor. However, I’m quite glad that we do not have to regulate ourselves but are given quite clear guidance by the GDPR on how we should treat sensitive data. We know what we can and can’t do. To me, that’s liberating.
Matthias (AiSight): The data aspect is interesting. Another point to think about is how Germany with its huge levels of industry still has vast amounts of machine data, which have not yet been used. Therefore, it is an advantage to be part of an AI company in the industrial space which focuses on this industrial data, because getting access to similar datasets to work with would hardly be possible anywhere else.
Tobias (whoelse.ai): Honestly, these are all good points but I am not sure working in Germany is per se different from working in other places in the world. Probably it’s even the opposite of the assumption inherent in the question. Germany is maybe a little bit like everywhere in the world, in particular when it comes to AI.
The German AI scene is powered by international talent. It is very common that teams of 10 people consist of 10 different nationalities. As a consequence AI in Germany is very intercultural. And this is great if we consider that a domain like AI explainability is subjected both to technical and moral questions. The diversity of people working in the German AI ecosystem, therefore, should give us hope that we will come up with the right questions about the future of AI.
Tobias (Whoelse): The one thing that makes Germany stand out in Europe is its decentralized ecosystem. We have fantastic research clusters in all parts of the country. That’s an obvious advantage.
Matthias (AiSight): I agree! There are fantastic research institutions and universities educating brilliant engineers. They are equally qualified to US engineers, but salaries are significantly lower. Therefore, it is possible for a German company to achieve more than a US company with the same investment, where machine learning engineers can easily earn 250k dollars per year.
I’d also argue that in some areas of AI, Germany is the leading nation in the world. One example of this would be like the Chair of High Performance Humanoid Technologies (H²T) at the Institute of Anthropomatics and Robotics, which researches and develops humanoid robotic technologies and systems that perform versatile tasks in the real world in interaction with humans. They count amongst the most advanced technologies in this field.
Matthias (AiSight): I believe that Germany is still very risk averse. For example, if you purely look at the money side: The investments — especially for later stage rounds — are still too low to compete with US companies. I mean, so far only 1.2 billion euros of venture capital has been invested in German start-ups that identify artificial intelligence as a core element of their business model. This sounds like a lot until you compare it with the American figure: The US has already invested 23 billion euros in AI start-ups over the last five years.
This means that ultimately, German founders will have to decide at some point whether they want to be funded by US investors and will have to adjust their strategy. It also means that plenty of great German AI talent is going to the US and Switzerland, because the salaries offered in these places are enormous.
Achim (Vialytics): And there’s also another frustrating aspect when it comes to the money side of things. I find it quite astonishing that when it comes to data, most data for training a machine learning model is only free for academic use. That doesn’t seem right to me. Commercial use should be allowed on datasets that are created by universities and are thereby funded by the state with tax money that mostly comes from businesses. While big corporates can invest in their own datasets or can negotiate deals with universities we as a startup face a disadvantage here.
Matthias (AiSight): There are obvious solutions here — it isn’t rocket science. Germany should invest more into research, universities and provide subsidies for AI projects in Germany. After all, while the German government plans to invest three billion euros in the development of AI by 2025, this is too little. Think about China: The city of Tianjin alone is planning AI funding of 12.8 billion euros. The Chinese company Alibaba has even planned to invest up to 16 billion euros.
I also think that the development of new technologies and their application should not be shaped by a few dominant economic and political actors, but should include voices from the economic, societal, and political spheres, and from every region of the world.
One way of achieving this should be a global AI governance architecture, which is politically neutral, international and interdisciplinary hub for basic AI research that is dedicated to the responsible, inclusive and peaceful development and use of AI. Germany could serve as host state of this international AI research hub and its implementation.
Tobias (whoelse.ai): I agree that research is vital: AI is a research-driven domain. It is not called “training” a machine learning model without reason. This paradigm of scientific trial and error can be a huge problem for the German industry which is used to watershed planning and project management.
But because the startup ecosystem is dependent on investments from these traditional industries, and large corporations in the innovation of small high-tech companies, it is important to improve the methodologies with which both can partner up.
Ultimately, I’d argue that Germany’s long-term strategy should be to become a leading integrator of different types of AI technology from all parts of the world. We are, after all, a manufacturing-driven economy. After initially combining excellence in electrical engineering and robotics (e.g. automotive manufacturing), the foundations for the next generation of business cases are now computer science. Germany needs to adapt to this change faster.
Want more hype-free takes on the AI scene? Bundesverband Deutsche Startups and hubraum have collaborated on a study about AI to deliver an overview of the current use cases, potential and challenges for AI in the German startup ecosystem. The study will also compare Germany’s innovation to Israel, in both the areas outlined above and in terms of digital ethics. Follow the upcoming study release on LinkedIn and Twitter.