MU's multi-graduate directs research in the field of verification at a prestigious German university

Jan Křetínský

We all see the development of AI in industry, which brings with it the need to ensure the quality of AI-based systems. This cannot be done with machine learning techniques, but also requires classical theory, including verification.

Author: Klára Petrovičová for

Doc. Dr. rer. nat. RNDr. Mgr. Bc. Jan Křetínský, Ph.D. became famous at the Faculty of Informatics of MU for the number of graduated universities and degrees he obtained. Behind these degrees are a number of successful publications and research. Jan Křetínský now works at the Technical University of Munich, where he teaches courses in theoretical computer science such as Verification, Automata Theory and Formal Languages, Complexity, Model Checking and many more. In addition, he leads a research team and is actively involved in research in theoretical computer science. You can read about what his studies were like, what he is specifically working on now and what he thinks is the difference between the universities in Brno and Munich in an interview for FI MU.

Why did you decide to study at three universities at the same time and how did you choose them?

There were basically two reasons. The first was that I wasn't quite sure what I wanted to do after high school and I didn't want to miss out on any of the opportunities that seemed attractive at the time. And the second was to have a broader perspective. Today, I often find myself appealing to interdisciplinary research, and having a broader overview of both related and completely distant fields is quite useful. When I made a particular choice, I was influenced by the situation in Brno and what would be easiest for me.

Finally, my choice was also influenced by the range I saw in the scientific hero Noam Chomsky, who also influenced the foundations of the emerging field of computer science, and it was Noam Chomsky who had a background in these areas. A lot of his results wouldn't have happened if he hadn't had a sufficient mathematical conception of the world and thought only as a linguist, and conversely, if he had thought only as a mathematician, he probably wouldn't have come up with grammars at that time. So it was also influenced by the fact that I saw that such a combination was to the benefit of the cause.

So even at the time of choosing a college, you were already familiar with these areas?

Even then in high school, I was reading books in linguistics, philosophy, and paradoxically, math and computer science, which I knew almost the least about, even though it runs in my family. (Laughs) I knew that I would get into computer science and mathematics eventually anyway, because they are sciences needed in all fields and there is a future in them, but I wanted to learn about the other fields I was interested in in high school.

And how did you get into that in high school?

One of the reasons was that I had a good high school that had a wide variance. There are also specialized math high schools in Brno that are high quality, but again, a lot of the other subjects there are quite underrated. I went to a grammar school that wanted to be really general, which I think was quite successful, and I had a lot of inspiring professors especially in the humanities, so actually through that.

So how did you end up getting into computer science?

I saw a lot of development and a future in computer science. It was clear to me that good computer scientists would be needed, but how many good philosophers does Europe need? Two? Three? (laughs)

How did you pursue your studies?

It was difficult. (laughs) Of course it wasn't free and I had to sacrifice a certain part of my personal life for it. But I have to say that at that time Masaryk University showed itself in the best light, because although it was not quite common at that time, it already offered possibilities within the information system to do all things easily, to combine everything, for example, so that I could easily have credits from different subjects recognized across faculties. I wouldn't say I had to put in three times the effort that I did at one school, but let's just say two and a bit times.

Then you decided not to continue in the philosophy department?

I saw that that was enough to get the experience, and further on you had to go more in depth.

In computer science you were involved in the area of formal methods and verification, why did you choose theoretical computer science?

Like a lot of decisions, it wasn't so much motivated by the substance as by the people who worked in the field. During my studies I came into contact with Professor Kučera and the collaboration came very naturally, if someone else had come along at that moment I might be doing something very different today.

How did your scientific career go after that?

Already during my master's studies, my family pushed me to look abroad, saying it was a valuable experience. I honestly didn't really want to go. (laughs)

I always liked to travel, but I didn't really want to leave everything at home and go out into the world without anyone I knew. So I limited it to Erasmus in Aalborg, Denmark, which was an excellent experience, in my case scientifically, because I was lucky again to have people who took me on, dedicated themselves to me and pushed me further in my career.

A big turning point for me was going to postgraduate. Everyone told me then that if you want to make it anywhere, you need Western experience. That's when it was confirmed to me that FI can compare with these universities in both studies and research. Still, it's something that won't be overlooked on a CV, and you gain valuable contacts for the future.

At the end of the day, scientific research is quite often about who you know and who you collaborate with, because it gives you access to something else and makes you a member of a community in a very different way than being isolated somewhere. So it's a bit unfair in the end, but it's the human factor that's always going to be there in some way.

So I did my PhD at TUM in Munich, but also in Brno. At first the vision was that there would be one under double supervision, which turned out to be so administratively complicated that it was easier to write two dissertations. (laughs)

After finishing my PhD, I went to the IST (Institute of Science and Technology) in Austria, where I was familiar with the environment because I had already worked with researchers there, so it was a very natural choice. And it was a good choice in terms of both the quality of the research and the breadth of interdisciplinarity. Even given that it's a small institution, interdisciplinary research and communication can take place more easily than at a large institution like Oxford. At IST, I had the opportunity to work with Krishnend Chatterjee and Thomas Henzinger, two huge figures in the field. They taught me not only to hone the scientific side of things, but also to work on the presentation side of things and also to know what to pursue, what not to pursue and what might be important.

Then there was an opening in Munich, which I eventually got to and I've been here for a number of years now, the experience is huge and the opportunities you have at TUM (Technische Universität München) are greater than many other places.

What do you think is the key to success in theoretical computer science? Is it mainly the aforementioned connections?

Two things are absolutely indispensable. The first is a basic education in computer science at the undergraduate level, because otherwise you miss the train. You can catch up, but it's hard. FI, for example, provides that at a world-class level. The second thing is effort and how much time you want to devote to research. Quite often I see a transformation in graduate students, from people who don't have much idea what they're doing and you wouldn't think they'd be great scientists, and you end up seeing them grow into new people in a couple of years and it's just that they've given a lot.

Then in the activity itself, it's important to do something that you're interested in, not just something that somehow floats your boat, and also to be a member of the community. Because at the end of the day, you will be judged by the people who belong to that community. The moment you're known in the community through your results, that's a huge help, and one way to be known for your results is to work with people who are already known for their results. So your results will definitely get noticed. Because quite often the overproduction of scientific papers is such that even the good ones don't get noticed, so that's the key part about the recognition. It's not that your acquaintances are going to push you somewhere, it's more about getting your work noticed by other people, and then it's always about the work itself. But sometimes it takes luck to get started. Also, if you get a lot of results with people you know, other people will say they probably don't credit you because you bring them coffee, but there's probably something to that.

What are you currently working on at TUM?

I'm working on verification, which is checking that systems do what they're supposed to do, and optimizing their reliability and availability. These days, it's typically systems that are "cyber-physical", so they have a physical component and need to work well. Because if these systems are already interfering with the human world and could hurt someone, that matters a lot more than just building an application which crashes all the time. These are the systems that have a lot of quantum in them, like probabilistic or temporal behavior that is not common in traditional software. So very often I work on analyzing probabilistic or temporal systems. I do both theory and, for me, interesting applications in the areas of robotics, artificial intelligence and biology. Not that I'm keen on the results of my work being immediately usable, but I find it good when the work goes in that direction. And I find it a terrible shame when one sees a huge advance in theory, but then looks at the practice in industry and finds that much of it is not reflected. It takes an awfully long time for industry to be able to use something on a day-to-day basis.

For example, we had a robotic arm in Munich that translated things in the warehouse and we tested the results on it in the area of explainability of robot controllers and Explainable AI, which is a big topic right now. In a situation where industry has a contract with a university, the link works, but for a company to go straight to an academic tool to do the work for them, that doesn't happen, you still have to have the university involved and that's not the target state.

How does the research itself take place? Is it just a pencil and a paper?

For the most part it's pencil paper, which is essential. But we're trying to do theory with the vision of making things practical, so we're not just interested in whether it's possible in principle, but we want it to be reasonable, so that the time and memory resources are reasonable. That is, it's not necessarily important that we don't get an answer in the worst case scenario, but if we give it a typical question that comes up a lot in practice and we get an answer quickly, we're happy and we see a shift, which is something that can help the industry significantly. Even though we know from theoretical results that we can't always get an answer quickly, we don't give up, but we try to do something. In these situations, you also need to have implementations and computational resources, so that's where having a larger number of students comes in handy, and having people who are immediately able to see if our idea is practically sensible or not and what path to take.

How big is your research team?

Right now we're around eight people. Then there are undergraduate and graduate students who join during their studies. This is one of the things I brought from FI, because here at TUM it is not so common for undergraduates to take a course called Guided Research, it is only for masters and undergraduates rarely join. This is something that FI can boast about globally. The care for students is very good in this.

Which area do you think will move from research to commercial in the next 10-20 years?

We all see the development of AI in industry, which brings with it the need to ensure the quality of AI-based systems. This cannot be done with machine learning techniques, but also requires classical theory, including verification. The other area, called Explainable AI, is one that almost no one knows how to deal with because it is extremely vague at the conceptual level. We may have European Union regulations on what all has to be met, but even the global leaders in the field are groping for what it all means. This area needs to link up with universities, because the industry is not yet able to orient itself in this direction. We need to come to at least some basic understanding in this respect before we can create concrete applications.

There is also a lot of talk about autonomous driving. No one would want to get into a car that is not safe. But the verification of these systems at a guaranteed level of safety is still in the future.

So these are the areas related to my research that are in demand in the marketplace and that must necessarily be developed in universities and hopefully, in the next generation, we will be able to provide something to the commercial sphere.

As far as the applied sphere is concerned, there has been close cooperation with industry at TUM for many years and whenever something is developed, there is an immediate effort to use it. There are world leaders in graphics, for example, where their results are directly used in animated films, for example, or in feature films, where they generate forests or seas, for example, or Deepfake video in real time.

Did you take anything away from Brno that you can incorporate into your teaching or research in Munich?

One of my important contributions is to integrate young students who start with me at the undergraduate level and go on to the PhD.

Another of my under-appreciated qualities is the information system. Even when I came here as a student and occasionally suggested to someone that we could have something similar to FI, everyone looked at me like there was no way we could do anything like that, even though it had been working like that in Brno for many years.

Of course, another memorable thing is the knowledge I got in my undergraduate studies, which is something I'm incredibly grateful for. Even the students who go from FI to Erasmus show that they have a good foundation.

What do you see as the difference between the universities in Brno and Munich?

One of the key differences is the ability to connect with industry. For basic research it may not play such a big role directly, but collaboration will bring more money and the opportunity to hire more quality capacity and be visible on a global scale. The university is growing dramatically as a result, here in computer science for example. They are all leaders in their field and they come here because of the reputation and the conditions. That's also created by money. No public institution can finance everything with public money alone. That's why TUM is ranked as one of the best institutions in the world.

Another thing that's being promoted a lot here is the establishment of student spin-offs. A lot of people are employed here just to help students start companies.

As far as education goes, I wouldn't underestimate Masaryk University. At least not in computer science.

On the other hand, what would inspire you from TUM if you returned to Brno?

It is a number of specific details that could be implemented. German education encourages practical courses, which you can't avoid, there is also more direction for students to go on internships. Practical and research projects are ECTS credit worthy. There is also more involvement of students in projects where they produce something within six months. I don't think the course-exam model is not good, but it shouldn't be the only part of education.

On the contrary, many things work significantly worse here, such as extreme benevolence to students who fail the exam. In Brno there is a system - three times repeated and then in a year, then it's over. There is nothing like that here. There are cases where you come to the exam after the 16th time and it still goes nowhere. You're always correcting the same exams over and over again. (laughs)

Where do you see yourself in 5 years?

I've always believed that you should do something that you see a purpose in. I'll see what comes my way, because a lot of the applications I'm doing have come about by accident. What comes along can be from all different corners. I always have a couple of projects on my desk that I want to take somewhere and then decide if it's worth continuing or starting something new.

Finally, what advice would you give to students who would like to pursue research in computer science?

I would definitely advise young students to get a sniff of research as early as possible, and if they see that a professor has a group of young people around, it's worth trying to ask around. And if you're scared, ask a member of the team if they could ask the professor for you. The earlier one goes in, the better starting position one has. After all, research is one of the most fun activities at university.

The female population generally has a harder time in computer science. For example, they often see that the male part of the population goes for it hard - they go to a professor and tell him that they would like to work with him on something. Women often don't have the courage to do that. As a professor here, I often try to address that and personally challenge students who might not have the courage to go for it. And then I have a gender-balanced group.

I would like to wish FI to continue in the direction she has set because I think she is well on her way. I also see a great effort of self-reflection and self-improvement. I would also wish Brno, as a city with great scientific potential, to continue to grow and improve in this direction, because I think it can go very far.

On March 29, 2022, Jan Křetínský presented his research work entitled A story of checking LTL models at FI for the professional public as part of his professorship proceedings.
Jan Křetínský

Gallery of graduates