Emeritus professor Arno Siebes has no plans to stop

Although he reached retirement age last year, Arno Siebes is still fully engaged in research and supervising PhD candidates. The Professor of Algorithmic Data Analysis simply enjoys his field far too much to call it a day. “Emeritus, as far as I’m concerned, is Latin for volunteer. I’m just carrying on, only the university no longer pays me a salary.”

Very much in keeping with his style, Arno Siebes formally stepped down last September after 25 years as a professor. With a symposium and a ‘fairly complicated’ farewell lecture on statistical patterns, he officially said goodbye to his field of Algorithmic Data Analysis.

The highlight of the day? Siebes doesn’t hesitate: the speech by two of his PhD candidates, “a talk full of praise”, ending with a Festschrift, an academic tribute volume filled with scholarly contributions and personal stories. “I hadn’t expected to receive one, and certainly not one published by Springer. When I saw how many leading figures in the field had contributed, I was genuinely speechless for a moment. And that doesn’t happen often.”

Doktorurgroβvater

Despite his official departure, the professor has no intention whatsoever of stopping; he simply enjoys his work too much. Since becoming emeritus, Siebes can still be found almost every day in his office on the fourth floor of the Buys Ballot Building. He continues to supervise PhD candidates, attend conferences and work on his own research. “I want to keep contributing to my field. The proverbial icing on the cake at my farewell was an overview of my academic descendants, in other words all the PhD candidates I’ve supervised. In German they call a doctoral supervisor a Doktorvater, so I’m now a Doktorgroβvater and even a Doktorurgroβvater. Brilliant, isn’t it?”

Arno Siebes started 25 years ago as Professor of Large-Scale Distributed Databases, although he never actually worked in that area. “I studied Mathematics in Utrecht and did my PhD in Amsterdam at the Centrum Wiskunde & Informatica, where I then worked for fifteen years on database management systems and later on data mining. When I applied for the job here in Utrecht, I said: I work on data mining, not data systems. They were fine with that. When the Faculty of Science was formed in 2005, along with the Department of Information and Computing Sciences, I renamed my chair Algorithmic Data Analysis and was able to shape our research field entirely myself.”

Nappies and beer

Siebes describes himself as a ‘data miner’, someone who digs through large amounts of data to uncover and analyse certain, sometimes unexpected, connections. This falls under unsupervised learning, meaning you feed a computer model with data and the algorithm then looks for patterns on its own. The model is given no hints about what is right or wrong. Siebes: “The standard example often used in my field is that analysis of large quantities of supermarket data shows that people who buy nappies also buy beer. That unexpected pattern led to the suggestion that beer and nappies should be placed near each other, or offered together at a discount.”

It’s a funny example, but is it really a pattern that keeps appearing, or just coincidence? Siebes: “The question I’ve been working on for the past fifteen years is how certain you can be that the model the algorithm uses to recognise patterns is actually the right one. There may be incorrect assumptions built into the model, or the data may be incomplete, which means the algorithm’s answer doesn’t match reality. My aim is to quantify that question, to come up with a way of calculating how reliable the model being used actually is.”

One of the crowning achievements of his career is that Siebes believes he has found an answer. It’s a universal solution that allows you to calculate how reliable the outcome is for any model, even if you know nothing at all about the data.

But he will never be completely certain, because unsupervised learning always involves a degree of uncertainty, unlike supervised learning. “My research is mainly a theoretical exercise, where the reasoning behind why I think it should be calculated in this way is what matters. There’s no formula you can simply plug things into; it’s uncomputable.”

Researchers are often driven by the simple wish to understand something more fully

Although Siebes calls this part of his research a “purely curiosity-driven hobby project”, its potential applications are certainly relevant to society. “The nappies and beer example is amusing, of course, but it doesn’t exactly change the world. For doctors, though, it’s just as important to know whether the models they use to treat or diagnose patients are reliable.”

Siebes says his own motivation is less lofty. “It’s nice to say you’re doing something for society, but for me it’s about science itself; as a scientist, I would find it deeply unsatisfying to know that my model wasn’t close to the truth. Researchers are just as often driven by the simple wish to understand something more fully.”

Although his field falls under what is often labelled artificial intelligence, Siebes prefers to avoid the term. “I find the term artificial intelligence rather pretentious. Especially those large language models, that’s not intelligence. They’re probabilistic models of language, and language and thinking don’t have all that much to do with each other; they happen in different parts of the brain. I don’t believe in books, films or music written by AI either. At best, it will all be fairly middle of the road.”

The AI bubble will burst

According to Siebes, companies currently pouring hundreds of billions into AI won’t earn that money back. “I see artificial intelligence as a toolbox of tools. One AI is good at predicting the weather, another always wins at the game of Go, and a third can rewrite texts. None of those tools on its own is the solution to everything. The hype is an AI bubble that’s going to burst spectacularly.”

However AI and data mining develop in the coming years, Siebes will keep following them as he always has: curious, independent-minded and with a healthy sense of humour. “Emeritus, as far as I’m concerned, is Latin for volunteer. I’m just carrying on, only the university no longer pays me a salary.”