Meet Jakub Šimko

The Gamechangers series presents the stories of the people who stood at the very beginning of the Kempelen Institute of Intelligent Technologies. Through their vision and determination, they transformed a bold idea into reality – creating a new place for excellent science connected with innovation and talent development in Slovakia. In the interviews, you will learn about the challenges they had to overcome, the values that unite them, and what motivates them to keep pushing the boundaries of research and innovation.

Jakub Šimko, whom you’ll meet in today’s episode, was one of the co-founders of the Kempelen Institute. He studied and worked at the Slovak University of Technology, has been involved in several European projects, and enjoys exploring ways to make technology fairer and more useful for people. For the past five years, Jakub has been one of the lead researchers at KInIT, focusing on algorithm auditing, machine learning, and the fight against disinformation.

The biggest challenge was figuring out where and how we would secure funding. We had some runway from our partners, which was generous and, one could say, unprecedented. But we knew that we would have to raise research funding in other ways. We also had a rough idea of how to do it – we wanted to pursue grants and industry collaborations. At the beginning, however, we didn’t yet know how successful we would be.

So for me personally, this was a very important question – how to finance it in the long term, what to do, which projects to take on and which not to… It was definitely a big challenge. But I should add that I didn’t really feel much pressure in this regard. The initial funding package we received to get started was sufficient for us to try out different things.

No, it wasn’t really stressful. Rather, it was a challenge that we threw ourselves into. For me, it was something new, because I hadn’t been used to writing grants on such a scale and in such depth before. But at KInIT I found myself mostly on the grant side and, paradoxically, we never had as much space for it as we did there. Since we didn’t have any grants at the beginning, we actually had the time to write them, because we weren’t burdened with too many ongoing projects that needed to be handled.

And in that initial stage, we also learned a lot about the process. It was probably during the first two years of our operation that we wrote and gradually secured a really substantial package of major European projects.

It felt like we were trying something new.

And what were those first months like? Let me put it this way – there were already about 25 of us, and it was necessary to get things organised. I remember the selection process for our first COO. Then we set up research teams and started introducing the first processes – how those teams would meet, where they would have a shared virtual space, what and how they would keep records, and so on.

All of this was happening during the COVID times. That meant we met very little in person and a lot online, which in some sense was actually good, since the whole world had moved into the online space, and it became the default mode of collaboration. Thanks to that, perhaps even the grant writing went more smoothly, because people got used to working from home and could fit in more meetings. And not just us, but our international partners as well… So we were able to move forward quite quickly.

The call was open that autumn, right when KInIT was founded. So, practically the very next day after we started, we could begin working on this project. There weren’t any other calls at that time, and it was the only major European project we could apply for then. So we threw ourselves into it with an unprecedented amount of effort. It was Majka Bieliková, Mišo Kompan, Robo Móro, Ivan Srba, and I working on this one project – our future CEDMO. And we had nothing else on our plates, so it was a rather intense month of work on this project. But that also made sense, because we weren’t very experienced with this type of call. For example, we were extremely meticulous in reading all the guidelines to make sure we fulfilled them properly.

Another new experience was the consortium, since we weren’t submitting the project on our own. We managed to become part of that consortium and, although we weren’t the project coordinators, I dare say we contributed quite significantly to ensuring the project was successfully created, submitted, and ultimately secured. That was a huge success and a real boost – it was truly a very powerful moment.

After CEDMO, there were no other open calls. So we focused more on the research itself, which until then had been quite gradual and mostly involved wrapping up unfinished work. But now we had the time, so we tackled the topic of auditing social media recommendation systems. Our first study took longer than we expected, but eventually we produced a paper that we submitted to RecSys 2021. And that was the first time, at least for me, that a full paper of ours was accepted at such a top-tier conference. First try, first hit – and we even received the best paper award… I thought, “Wow, amazing, this is what it looks like when you really throw yourself into something seriously.”

And with that, I would perhaps conclude this answer: my most important insight from that first half-year was that when you truly focus on something at KInIT, the results can be far more impactful than under other circumstances.

It often has a social dimension. We work on processing content from the web and social media, and on enabling machines to understand that content. Of course, this can be applied in many ways. When the topic of false information started emerging around 2014, it seemed to us that it would become important.

For me personally, it’s about the fact that I’ve been working on this topic for ten years, and after that amount of time, you naturally start to become somewhat of an expert. The same goes for my colleagues. It’s a field with plenty to work on, so we continue to engage with it.

Yes… In 2014, very few people saw it as an important issue – the spread of disinformation and propaganda, or the fragmentation of society into bubbles and the related online radicalisation. And this is directly connected to social media algorithms, including the use of networks for political purposes, and further fragmentation and radicalisation. Then came 2016, and the issue gained more attention, but public institutions still weren’t really addressing it. In a sense, that’s still somewhat true today, although, for example, the European Commission has begun to recognise it as a threat to the democratic system we have.

So even though more and more people started to engage with the issue, the situation worsened in the meantime. Malicious actors began pushing hate more aggressively, learned to exploit the media space to influence social media users, and gradually, their online presence intensified.

This escalation is extremely slow, but it’s there. So even someone who ten years ago would talk to everyone and wasn’t particularly opinionated can gradually become radicalised if these narratives are dripped to them over a decade. That person becomes at least crypto-radicalised. This means they would never take any physical action, but they start expressing extreme views – and of course, they won’t think of themselves as being extreme at all.

But I’ll also say that for us, this has always been just a domain where we apply methods of automated text processing, machine learning, and artificial intelligence. So we’ve always been only one part of the picture. We’re not interested in the specific narratives or actions themselves – that work is handled by others: fact-checkers, strategic communicators, and so on. Professionally, what matters to us is whether we can understand automatic text or online content well enough to support these people and make their work easier. That has always been our main goal.

Research and curiosity are extremely important – they simply have to be there. And everyone should also be mindful of social responsibility, because we are trying to build something that goes beyond ourselves. Something that is meant to have lasting value and to elevate our country.

We must not let our guard down when it comes to sustainability. I see it optimistically in the sense that over these five years we’ve managed to extend our runway even further – and that is already a certain measure of success.

So, another major “we did it” moment was securing the lorAI project. That’s a huge deal. Thanks to it, we now have the opportunity to consolidate, and for a much longer period. At the same time, it allows us to better set up processes and manage growth, which is also a big challenge.

Let’s be honest – once we have 100 members, we need to organise things differently. We are growing continuously, and some procedures cease to be adequate. And usually, you only start addressing something when it begins to hurt – and then you end up tackling many things at once. So the biggest challenge ahead of us is how to manage organisational changes in a way that allows our institute to function even larger than it originally was.