"AI will change our world even faster"
InnoTrans Report: Mr Böhm, on which exciting product is your com- pany currently working?
Thomas Böhm: It is difficult for me to pick out one, because all our products are exciting and each for a different reason. On the one hand, we have KONUX Traffic, our latest product, which uses data fusion and AI to create a historical model of capacity utilisation and delay events. It is designed to reveal reserves and increase resilience. For the first time we are mapping the train traffic of an entire network. It excites me to see what is being created in such a short time. For example, when I saw for the first time how train delays spread, "move" and disappear across the network over a day. The other is KONUX Network, with which we bring together actual load, conditions and external events to predict ageing and maintenance needs. For the first time, we are addressing the major challenge of linking data and, above all, analytical models across several products and very different use cases. We're racking our brains on how to depict logical relationships while making efficient use of cloud infrastructure.
And then there is the predictive maintenance of switches, which is currently being rolled out at Deutsche Bahn. Here, it is exciting to face surprises which we have never thought of before, for example, a situation in which the mobile network in an entire region is unavailable for more than a day. At the same time, we are continuing to develop the system, for example for monitoring switch blades.
Where do you see the biggest changes through AI?
Thomas Böhm: In the handling of large unstructured volumes of data, in a boost in productivity and in interaction with computers in general. AI has been part of KONUX for a very long time and for me, as a trained data scientist, it has been part of my life for even longer. Therefore, only the breakthrough of generative AI was surprising for me. And this is exactly where I am experiencing significant changes at KONUX, in my environment and in the "bubble" in which I find myself. In our company, AI has recently been helping us to write software. And some colleagues are programming again thanks to ChatGPT or CoPilot, because they can now again achieve results in the little time they have.
In what ways can AI put the rail sector on track for the future?
Thomas Böhm: The answer to this question is directly related to the previous one. Many infrastructure and transport enterprises as well as manufacturers and suppliers will be relieved of administrative work. This will make the sector more agile and cost-effective. Furthermore, AI seems less alienating or mystical, and I believe that AI-driven products will be accepted more quickly as a result. There are already numerous solutions that use AI to solve problems in the rail sector, for example InstaDeep in the dispatching and control of rail operations, Machines with Vision in tracking and the AI solutions of Deutsche Bahn itself. The biggest lever, however, is the ability to interact with computer systems using natural language in generative AI. This gives more people access to problem-solving methods - such as machine learning - and solves more problems which today require an in-depth analysis.
What areas does AI not have an impact on?
Thomas Böhm: I don't think there will be anything that AI doesn't have an impact on - just as there was actually nothing that the internet didn't have an impact on. But there will be some areas, namely in the rail sector, which will be less affected, such as all kinds of work with a desirable, direct contact to humans. Likewise, physical work such as maintenance, construction and assembly will be less affected, as will anything involving direct contact between people. I hope that increased productivity in administration will result in increased salaries and attractiveness in maintenance. As soon as I suddenly don't have to document everything three times, there will be more time for repairs.
What are the growing challenges which the use of AI will bring for transport companies?
Thomas Böhm: AI will change our world even faster, which will put even more pressure on transport companies to be more modern and offer higher quality and more attractive jobs. In addition - and this applies to all companies - we will also have to demand much more transparency and prove authenticity. Many AI systems are supposed to get better over time and adapt to new circumstances. But AI also makes mistakes and can even degenerate. That's why we already implement a monitoring system for our models and systems during their development.
What decisive developments in the use of AI have taken place in the railway sector since KONUX was founded almost ten years ago?
Thomas Böhm: AI is no longer a fancy topic, no longer an abstract future. Many fields of AI have arrived in use - even if not yet across the board. In addition, many companies understand better that opening and sharing data dumps offers a lot of benefits. It is simply better to work with them than to hide data.
Is KONUX already working on ideas which the industry has not yet thought of?
Thomas Böhm: That's almost an unfair question, because I can't know what all the other clever colleagues in the industry are dreaming about. The best I could imagine is that we are already much further ahead in how we keep the human being in the loop. Especially in the areas with very specialised and implicit knowledge, AI systems will take a little longer to learn. And this will also be the case where the relevant events, such as an infrastructure disruption, are rather rare. That's where we need humans and their experience to improve systems. That's why we are also working on integrating them.