Thomas Thiele: AI is Team Sport

Deutsche Bahn carries people and goods from A to B. As banal as this may sound, it is ideal for understanding artificial intelligence (AI) even better and, above all, for grasping its social significance. This is why we spoke with Thomas Thiele, the man who is driving AI for Deutsche Bahn and is also a member of the advisory board of Peter Feldmann’s AI Frankfurt Rhein-Main e.V. association. The mobility and logistics giant has probably recognized the urgency of optimizing classic railroad technology with new technologies.

German version: here

Dr Thomas Thiele is a mechanical engineer. He studied at the RWTH Aachen and received his doctorate. When he talks about AI, one can guess from his words where the man comes from: from application-oriented research. He has now left this domain behind him. He wants to make railroad technology smart at Deutsche Bahn. He says that it still has components that are 180 years old — even though the actual components are of course newer.

Thiele is a pragmatist through and through: “The reading is that AI is a tool for DB. We are not Google or Amazon. We are a corporation that transports people and goods from one place to another, a heavy metal corporation, to put it bluntly”. And AI is just a tool that enables the railroads to fulfil their core tasks even better. But this is not due to any play instinct. It’s solely due to the fact that traffic has increased significantly so far. “In recent years, we have seen a steady increase in the number of passengers on long-distance services and we expect the figures to rise again in the future,” he says, describing the situation on the railways.

And all this is what Thiele describes as the central motivation for his actions as an AI expert together with the DB Management Board. Incidentally, Thiele thinks that the whole issue is a social development with a simple crux. “We travel more and more. At the same time, ecological awareness is increasing. As the most climate-friendly means of transport, the railroads are of course benefiting from this, but at the same time, they have to create much more capacity very quickly.

So despite all the unpunctuality, rail is still an adequate option? Seems so. But how does Thiele intend to reconcile this with the railroad infrastructure, which is already very busy? It’s not the rail networks that can be expanded indefinitely, nor is it the number of vehicles that can be increased excessively. “Of course we are gradually expanding both. But at the end of the day, these are all things that take several years and decades, especially the expansion of the infrastructure is a long-term undertaking”.

And this sluggishness is now contrasted by something vibrant here, incorporated in artificial intelligence. AI, thank you, because “it promises us a short-term profit. The time scale on which we can realize it is definitely shorter.”

So we ask ourselves how this should look in practice — and are pleasantly surprised. Because Thiele unpacks a nice example. Namely the traveller app, the so-called DB Navigator, which we all know. It has been using machine learning processes, i.e. AI methods, for around two years now. A forecasting procedure works in the background to determine the punctuality of trains. It is only visible to passengers by means of a small delay display in the train app. Otherwise, this AI is not noticed at all. The norm-based forecast has now become an AI-based forecast method.

And we already thought that the railroads were about to use AI to forecast passenger numbers or trainload. Thiele laughs: “We have a very diverse portfolio of different use cases that we are currently implementing. But forecasting passenger numbers is like forecasting stock market prices. It would be great if that were possible, but that is also subject to social influences and therefore sometimes unforeseeable events, as we have noticed in recent months.

Arrived! Next topic: Value creation cycles

In addition, the focus on the car as a cultural component is currently washed out in Germany. The automotive industry is also in a state of upheaval. But stop: Is that why people are on the road less? No! And yet the railroads have to react to this somehow. And DB is doing that with AI. “And that is precisely why DB has already launched several AI projects,” says Thiele.

What else? In traffic planning, Deutsche Bahn is testing artificial intelligence methods, namely reinforcement learning. And where traffic has to be directed to specific situations, where decisions often have to be made in a short time, AI is the method of choice. Its procedures can then be used to put more capacity and vehicles on the rails. In this way, the Group is responding to the increasing complexity of traffic, which can be managed with new tools.

AI is also used to get these tools up and running by linking all the company’s possible processes. The integrated view is the top priority. “The vehicles that we want to get more on the rails with AI have to be available in the first place. So we also have to look at the issue of maintenance in this context. And here, too, AI procedures offer starting points for diagnosis and linking processes. So that we can say that data transmitted by the vehicles can be evaluated to predict their condition and also their component failures ex-ante.

Mr Thiele, it all smells as if AI is a team sport?

“That is very true! And because it is, we also want to participate in networks such as the AI Frankfurt Rhein-Main e.V. association, which was founded by the City of Frankfurt. Deutsche Bahn has a special position in the corporate landscape because we are state-owned. This means that the supply of mobility and logistics services is naturally very central for us. We want to be part of such networks in order to offer mobility that we develop together with the city. The association offers us interdisciplinary exchange beyond our own horizons. All of us, i.e. representatives from business, politics and research, would like to develop perspectives and, of course, jointly set new things in motion. I think that a city like Frankfurt can certainly serve as a testing ground for the live testing of new processes. I think this is very exciting and has prompted me to accept the call to join the advisory board.

Many thanks for the interview, Thomas Thiele!

Journalist // Blogger // Podcaster with focus on Artificial Intelligence and Data Science /// AI Series — SAS Hidden Insights