AI in occupational safety? But of course!

Scientists who are knowledgeable about artificial intelligence are in high demand: Martin Brenzke, for example. The physicist worked at the Jülich Research Centre for several years before the Federal Institute for Occupational Safety and Health in Dortmund was able to recruit him. As different as these institutions are, the research questions Brenzke pursues there still revolve around artificial intelligence.

Mr Brenzke, let's start at the beginning of your career. What were you working on at the research centre? 

I did my doctoral thesis at what was then IEK-4 – now IFN-1 – from 2018 to 2022 and continued to work there as a postdoc until 2024. The topic was fusion reactors, or more precisely, the question: Can the heat loads generated in fusion reactors be modelled using machine learning methods? So basically through AI, even if the term is of course broad. Using experimental data from the ASDEX Upgrade fusion experiment in Garching, I developed and adapted a model that allows us to map the physical processes. This is less time-consuming than the simulation codes that would otherwise be necessary.

The path from there does not necessarily lead to the Federal Institute for Occupational Safety and Health. Why are physicists needed there? In other words, what is your current role?

The connection lies in AI: I wanted to continue working with machine learning processes – albeit from a more critical perspective. I am by no means interested in jumping on the hype train and advocating for more AI in all areas. Rather, I believe that artificial intelligence should be used sensibly in a regulated manner so that it does not get out of hand. The Federal Institute has set up a new specialist group on the topic of ‘AI in the world of work’, which is how I ended up here as a research assistant. So I'm still sitting at my computer writing code. The idea is to look at AI from an occupational safety perspective and ensure that it is used sensibly in the workplace. Among other things, I am investigating the extent to which companies are already using AI – in my case for management tasks. Where are the potential risks for employees? And how can these risks be reduced or even eliminated using technical means? The first hurdle is to get companies to talk to us: I am currently looking for companies that will give us a rough idea of the tasks for which they use AI. If I cannot find any such companies, the risk factors will be assessed on the basis of theory.  I will then work on technical measures to eliminate these risk factors.

So you have remained loyal to AI. How does your current work differ from your work in Jülich?

The focus of the content is different. Whereas in Jülich the technical aspect was more in the foreground, I now also take a socio-technical perspective. The research methods also differ accordingly: for example, conducting interviews is new to me. I am also more involved in knowledge transfer and external communication, for example, I occasionally give lectures and workshops on understanding AI – not only at purely scientific conferences, but also to the interested public. Most recently, for example, as part of Digital Week Dortmund. Another new experience for me is participating in standardisation committees, such as at DIN.

Is there anything you miss about the research centre?

What I found cool about the research centre was the variety of topics and areas of focus at the institutes. Even though I wasn't able to benefit from this as much as I would have liked, as the coronavirus pandemic coincided with my time there – which meant that interaction was somewhat limited. The campus is also in a great location, surrounded by nature and deer that aren't shy at all. My current employer is in the middle of Dortmund, which is not quite as scenic.

Professionally, you spend most of your time working with AI. Is your free time also dominated by artificial intelligence?

Privately, I try to avoid AI. On the one hand, large corporations are behind ChatGPT and other widespread AI systems, and on the other hand, I don't see the benefit for myself privately. Every now and then I use DeepL for translations – at least there's a German company behind it. As for computers, I do spend a lot of time on my PC at home, but I use it more for gaming or music. I rarely tinker with AI algorithms in my spare time. I prefer to pick up my guitar: I recently joined a band, but we don't have any gigs lined up yet.