“Data analysis is a field of the future. But while American tech companies are already setting standards, Germany is lagging behind – partly because there is a lack of opportunities for students. What universities can do.
Data, data, data. It shapes our lives, replaces experience, and, for example, helps us find new ways to avoid construction sites or traffic jams. "Above all, it reveals connections that aren't immediately obvious, and that's what I find so exciting about it," says Hannah Feltz. The 23-year-old is studying Data Science at Hamburg University of Technology (TUHH). Even before starting her studies, she encountered the term data mining at the Chaos Computer Club – the analysis and extraction of insights from datasets using algorithms. "An underestimated field – and one of my favorite subjects in my studies," says Feltz.
In her fifth semester, the student took a mandatory data mining lecture and, in the accompanying seminar, was given a dataset of Hamburg's landmarks. Cleaning, clustering, coding – the students were supposed to demonstrate what they had learned so far in their studies and analyze the data professionally. "The hypothesis was that Hamburg is made up of several small city centers that have merged together," says Feltz, showing a city outline dotted with many colorful points. "You can't see much there because everything is on top of each other," the student comments on her own graphic and, in the next step, selects only the oldest eighth of the landmarks for viewing. The cluster ranges from the year 1263 to 1800, such a large time span because less data is available. Some spatial clusters can be recognized in principle, but there are many outliers, and the result is not conclusive.
It is precisely this openness to unexpected results that characterizes scientific research. "We didn't really find anything," says Feltz. Nevertheless, the group work was fun: "The tasks were open-ended, you could really let your creativity flow, which I thought was cool." Programming the clustering algorithms took the students about seven hours, so they were a bit exhausted by the time they got to the graphical presentation. Although the compelling visualization is essential for data mining, Feltz emphasizes: "A good plot says a lot, whether it's a bar chart or a line graph." The student would have liked to demonstrate this in her bachelor's thesis, but she didn't have a suitable data mining topic in mind—and neither did the professor she consulted.
Yet, data mining is a field with a promising future. Companies are spending a lot of money classifying their rapidly growing data sets or identifying patterns, outliers, and trends.
And not only companies: Ministries recently made headlines for purchasing police software from the US company Palantir.
From the perspective of data protection advocates, this was a mistake because it's unclear exactly how the data is linked and how profiles are created—and there are concerns about connections to American security agencies.
"We absolutely must develop such expertise and software ourselves to avoid dependence and remain competitive," demands data science professor Marina Tropmann-Frick. She is a data mining expert at the German Informatics Society. At the Hamburg University of Applied Sciences (HAW Hamburg), she conducts research for the German government on "Responsible AI."
Tropmann-Frick sees the European "AI Act" for regulating artificial intelligence as a positive development. "The whole world is watching us, amazed that we even dared to do this," says the professor. Europeans are good at tackling problems in a structured way, which ultimately leads to more robust software. And German universities could contribute to this: "Germany doesn't lack expertise, but it needs to be better supported," says the data scientist.
Government contracts and venture capitalists, which made Palantir so successful, are one thing.
More professorships, more funding programs, and more collaborations with medium-sized software companies, on the other hand, are the foundation for being able to develop such complex analytical tools in the first place.
Overall, Tropmann-Frick calls for more data mining courses at universities: "We may not be offering enough, and we also have a resource problem," she says.
The professor does not rule out the possibility of a dedicated degree program. She argues against this, saying it wouldn't make sense: "From data management and data preparation to data mining, there are fluid transitions; it all belongs together."
Regardless of whether a degree program is titled Data Science, Data Mining, or Machine Learning, it always involves broad research areas, including statistics, databases, and artificial intelligence. "Ultimately, all these fields are concerned with identifying patterns, relationships, and predictive models in the data," says Bernd Bischl, who primarily teaches machine learning in the "Statistics and Data Science" program at Ludwig Maximilian University in Munich. Bischl initially studied computer science. Then, in Edinburgh, he studied AI and machine learning – and had to push through the ML degree program against the advice of his trusted professor at the German Academic Scholarship Foundation. "In my case, it was a physicist who tried to talk me out of it because it wasn't considered proper science," Bischl recounts.
That was in 2001, when dot-com startups had already clearly demonstrated that data science has solid mathematical foundations and can be economically successful. "We missed an opportunity there," says Bischl.
Today, catching up with the lead across the Atlantic requires massive amounts of data, expensive computing power, and close collaboration with a tech company for the necessary infrastructure.
"We have several shortcomings simultaneously," the statistician says, then immediately corrects himself. He doesn't want to sound so negative: "Germany has really done a lot in the last 15 years. The six AI competence centers, in particular, are a big step in the right direction." One of them is the "Munich Center for Machine Learning," or MCML for short, in Munich – and Bischl is one of its four directors.
"20 million euros a year, that's no small sum, to bring together the best people in machine learning and AI in Munich," the professor praises.
The ability to extract empirical knowledge from data and use it to make rational decisions has become a key skill, especially since the COVID-19 pandemic. To achieve broader impact, the topic needs to be introduced in schools. More research and development of proprietary software requires young talent willing to dedicate five years to intensive study in data science, not just through continuing education. Europe-wide university collaborations expand the range of degree programs and foster international contacts.
Professor Tropmann-Frick recommends double-degree programs, where students earn degrees from two universities. "This can be financed through an Erasmus scholarship."
Afterward, many options await; data scientists are in high demand, and the application possibilities are vast. Tropmann-Frick has thus delved into particle physics and is researching a ChatGPT for particle physicists. For Hannah Feltz, it's federated learning, a machine learning technique in which a model is trained decentrally on multiple devices – and is now the subject of her bachelor's thesis. "That's not related to data mining, but it's also exciting," says the student. Afterward, she plans to pursue a Master's degree in Data Science at TUHH (Hamburg University of Technology), but what comes after that is still open. "Currently, I'm studying, and I'm really enjoying it," she emphasizes. Police IT is probably not in the cards, as it doesn't appeal to her, but a European version of Palantir is a possibility. The student is confident about this: "We have the expertise, here at TUHH as well, but it takes time and resources to write such software." [1]
What a brave student, alone ready to replace military-industrial complex of America. Those professors getting 20 mln. Euros a year for their own food expenses are living a good life. For machine learning research 20 mln. Euros a year is a funny small number though.
1. Ein eigenes Palantir? Frankfurter Allgemeine Zeitung; Frankfurt. 01 Nov 2025: 30. Von Deike Uhtenwoldt
Komentarų nėra:
Rašyti komentarą