Paul Mühlenhoff, head of the GDNÄ student program, on amazing experiences with ChatGPT, attempts at cheating at school and discussions in the staff room. 

Mr Mühlenhoff, there has been lively discussion about ChatGPT for months. Do you have your own experience with the text generator?
I tried ChatGPT for the first time in late autumn 2022 and found the program surprisingly good, even in the then still early version. I asked practice questions for an upper school lesson on the novel “Der Trafikant” – just for testing purposes, not for a concrete application. The results came very quickly, were quite challenging and seemed to be suitable for the target group. I found it impressive that the program told me what it didn’t know yet. Despite some shortcomings, it was already clear then where the journey was going: these systems will get better and better. 

In the debate about ChatGPT in schools, some call for a ban, others emphasise the new opportunities. What is your position?
In my opinion, a ban would be pointless. ChatGPT and other generative language models are coming and we have to deal with them. Our job as teachers is to explain how it works and to prevent abuse. 

According to a survey by the digital association Bitkom, half of the students in Germany have already used ChatGPT when doing homework, writing texts or preparing for exams. What is it like at your school?
At my school, the Helmholtz-Gymnasium in Bielefeld, ChatGPT is mainly used in the upper school. I don’t know exactly what the situation is like in the middle school. I suspect that it doesn’t play a big role there yet.

© Timo Voss, Studio of Thoughts | Helmholtz-Gymnasium Bielefeld

Bielefeld’s Helmholtz High School, shown here in an aerial photo, was founded in 1896. Under the motto “A modern high school with tradition”, around 100 teachers now teach around 1000 students.

For what purposes do the young people at your school use the chatbot?
As far as I can observe, mainly to try it out and play. However, there have already been individual cases of suspected misuse this year. For example, it was a question of subject papers, i.e. papers that are written at home without school supervision and whose grade is weighted the same as an exam grade. If there are clear discrepancies between the work and the previous performance, we naturally become suspicious.

How did your school react in this situation?
Since we have the burden of proof, colleagues scrutinised the work very closely. Knowing that ChatGPT can also generate source references, they checked, for example, whether Bielefeld libraries had the books mentioned in the paper. The suspicion that papers were written with the help of other people, such as parents, also existed in the past. But ChatGPT opens up completely new dimensions. For this reason, there is already a discussion in our staff about whether papers of the previous type will be acceptable at all in the future. Will they have to be supplemented by oral examinations or will we have to find completely new ways?

It will probably be the same in other schools. How are the school authorities reacting to the challenge?
The Ministry for Schools and Education in North Rhine-Westphalia reacted quickly and published a well-done guide for dealing with text-generating AI systems. Other federal states have also published corresponding recommendations.

What about teacher training on AI in general and chatbots in particular?
There are such offers. But they would only be interesting for me if they were specifically tailored to my subjects of German and biology. That is not yet the case.

You tested ChatGPT early on. Did that have consequences for your teaching?
Yes, I promptly made the new technology a topic in my upper school lessons. We talked about how chatbots work, what they can and cannot do. We also talked about the uncertain source situation and data protection concerns: After all, you have to give out your mobile number to use ChatGPT extensively. However, my main concern in the conversation with the students was to warn them early on about the temptation to use the AI model as a “homework helper”. Using it would be beneficial in the short term, but if you don’t think for yourself, you won’t learn properly. And that is the danger behind it. But then we didn’t continue to use ChatGPT in class.

Can you and your colleagues tell whether a homework assignment was done by students or by a chatbot?
So far I haven’t had any suspicious cases. I would argue that we teachers can tell pretty quickly whether it is our own performance or not. We can assess the performance of our students very well through their participation in class and the exams – discrepancies are quickly noticed. In my subjects, I hardly see any scope for using ChatGPT for homework anyway. They usually involve material with texts and graphics, and you can’t easily feed the AI with that yet. With simple definition tasks, it is perhaps something else. But these are not of great importance for us.

In what way could ChatGPT be useful in your subjects?
That’s a difficult question, because the areas of application are incredibly extensive and many questions are still unanswered for us teachers. I see a possible use especially where the students are in dialogue with the AI and have to formulate their prompts, i.e. their requests, more and more precisely in order to reach the goal. Learning then takes place in this process. In biology, for example, it would be conceivable to first develop experiments hypothetically or even to model an experimental procedure. In German, for example, I would find it exciting to have the AI interpret specific dialogues from a novel or drama excerpt and to discuss the plausibility of the reasons with the students. Can an AI grasp irony? How much context does it need for that? Answering these questions would also be useful for our own understanding of literary texts.

Will ChatGPT fundamentally change schools, as some predict?
No, I don’t think so, at least not in the classical school subjects. It might be different in computer science classes. What certainly remains unaffected by the new technology is our mission as teachers: we are supposed to educate young people to learn independently and impart knowledge. ChatGPT can possibly help with this and enrich the lessons, especially in the upper school.

© MIKA-fotografie | Berlin

Grand finale of the 2022 Student Program on the occasion of the 200th anniversary of the GDNÄ in Leipzig. Paul Mühlenhoff is is in the middle of the second row from the top.

Peter Liggesmeyer, computer scientist and member of the GDNÄ Board Council, on artificial intelligence, obstacles to innovation and affordable cell therapies. 

Professor Liggesmeyer, everyone is talking about ChatGPT these days. Is this also the case in professional circles, for example at your Fraunhofer Institute for Experimental Software Engineering IESE? 
Yes, chatbots have been talked about here too, and not just for a few months, but for years. At IESE there is a general interest in the topic; at the University of Kaiserslautern, where I have been teaching and researching for almost 20 years, it is about the use of ChatGPT in teaching or the evaluation of student performance. These issues are the subject of controversy among colleagues. But I think it will be clear that a complete ban on the use of systems like ChatGPT makes as little sense as its unrestricted use. 

What is your position?
Linguistic models such as ChatGPT can produce polished texts from a few keywords, but they are no substitute for the often tedious and laborious knowledge acquisition that university is all about. The models are suitable for quick bibliographic searches, for example, and can provide valuable services in this regard. The field is currently very volatile, there are good arguments for and against the use of chatbots in universities. I think we should observe the development for a while and make decisions after a reasonable period of time. 

ChatGPT has brought the topic of artificial intelligence sharply into the public consciousness. How do you see the development in this field?
I too am amazed at how fast everything is going and what is possible today. The results are getting better and better. They can no longer compare with the first bulky AI solutions that I worked on as a PhD student in the early 1990s. The big push came in recent years, mainly thanks to high-performance computers and the availability of large amounts of data. But, of course, even these impressive systems are not intelligent in the true sense of the word. 

Where does Germany rank in AI research in international comparison?
With the nationwide German Research Centre for Artificial Intelligence (DFKI) and its partner companies, we can keep up a good pace in research and development. In addition, there are numerous AI researchers at universities and research institutions, including Fraunhofer IESE. Overall, I see great opportunities for our country here. One obstacle to innovation could be the planned EU AI Act, a law to regulate AI applications at the European level. The aims of the project are quite honourable; however, I fear that the envisaged implementation might become an obstacle for the technology with negative effects also in practice. We know this from the General Data Protection Regulation (GDPR), which pursues very sensible goals, but confronts us daily with stuffing masks when using the internet in the name of online data protection. We should avoid this at all costs.

© Fraunhofer IESE

Workspace at Fraunhofer IESE.

As a researcher, you have been working for thirty years at the intersection of computer science and classical engineering. Is there a common thread between your projects?
Yes, there clearly is. At the beginning of my scientific career, I focused on software quality assurance. Now I focus on the security of digital systems. This involves two types of security risks: external threats to systems in the sense of the English term “security”, but also “safety”, i.e. dangers emanating from the systems themselves. Security risks exist, for example, in self-driving cars, in autonomous Industry 4.0 environments or in medical technology. If these systems are to make autonomous decisions with the help of machine learning, residual risks must be determined as a basis for the required certifications. This is not currently possible for machine learning components, but it would be important, which is why research is being carried out. Humans are good at constantly finding workable solutions for complicated tasks with imponderables in everyday life. The technical solutions of the future will have to be measured by whether they can do something similar. 

Do you have any examples of this?
Yes, specifically MY DATA Control Technologies from my Fraunhofer Institute. This is software for individual control of data usage. It allows you to define and monitor compliance with data usage rules. For example, one could authorise the use of one’s own data for medical research purposes, but at the same time set up a comprehensive data protection that prohibits the use of the same data for advertising purposes, for example. What happens to one’s own data is always determined by the donor of the data. The system arrives at workable compromises in complicated situations with simple rules, as we humans do every day. 

How important to you is the practical use of your research?
The practical orientation of my research is important to me. If the results are then used in practice, that is, of course, especially motivating. With this in mind, the artificial contradiction between basic and applied research dissolves. Industrial and academic research also come closer together. Thus, at my university I am involved in the areas of “Commercial Vehicle Technology”, “Works of the Future” and “Region and City”. At Fraunhofer-Gesellschaft, I am a member of the spokesperson team of the strategic research field “Intelligent Medicine”.  

Intelligent medicine: what does it mean?
Together with several Fraunhofer institutes, we are currently developing automated production technologies for new vaccines and mRNA-based cell therapies. We are using solutions that have emerged in the context of our Industry 4.0 research to be able to produce highly effective and affordable individualised medicines in the future. A very smart and sensible idea, in my opinion.

© Fraunhofer IESE

Main building of Fraunhofer IESE in Kaiserslautern.

You have been a member of the GDNÄ Board for a few months now. What motivated you to accept this honorary position?
On the one hand, the interdisciplinary nature of the GDNÄ, which fits in very well with my professional background. I graduated in electrical engineering in 1988 with a specialisation in data technology. The combination of computer science and engineering runs through my career. In the GDNÄ I would like to advance these subjects and, at the same time, promote their interdisciplinary linkage. For example, by building new bridges with the German Physical Society or the Society for Computer Science. There are also interesting points of reference between engineering, computer science and medicine, for example in the field of RNA therapies. I see the GDNÄ as a crystallisation nucleus for promising co-operations.  

Could this be an argument for young academics to get more involved in the GDNÄ?
I am convinced of it. Young people know how important interdisciplinary cooperation is for real progress. It is very important to be rooted in one’s own subject, that is something I want to point out. Because only those who are well versed in their field can be successful in interdisciplinary teams.

Heike Rauer, Director of the Institute of Planetary Research at the German Aerospace Centre in Berlin-Adlershof, on a new space telescope and her work with young people. 

Professor Rauer, your talk at the 200th anniversary of the GDNÄ was about the old human question of whether life is possible beyond Earth. A few months have passed since then. Have you come a little closer to finding the answer?
I think so. We are making good progress with the work on the PLATO space telescope, which is scheduled to launch at the end of 2026 and can detect Earth-like planets in the Milky Way. We hope PLATO will provide groundbreaking insights that will help us answer questions like these. 

You are part of the PLATO management team. How can we imagine this project?
It is a major scientific project launched in 2014 by the European Space Agency ESA, involving more than a hundred research institutions and the space industry. The acronym PLATO stands for PLAnetary Transits and Oscillations of stars. This mission will help us estimate how many Earth-like planets there actually are. We can then study the atmospheres of discovered planets with large telescopes such as the James Webb Space Telescope and its successor projects. From a distance of 1.5 million kilometres from Earth, PLATO will study star systems in the Milky Way. It records the brief obscurations that occur when planets enter the space between the star they are orbiting and the telescope. In addition, PLATO measures the seismic oscillations of the stars themselves. Once we look at this data collectively, we will not only be able to infer the mass and radius of the planets, but also determine their age – much more accurately than has been possible so far. 

How many planets are we talking about?
More than five thousand planets outside our solar system, so-called exoplanets, are known today. The nearest planetary system, Proxima Centauri, is 4.24 light years away. The most distant known exoplanets are 22.000 light years away from us. A journey to these planets would take thousands to millions of years with today’s technology and would therefore be impossible. But with space telescopes like PLATO, we can gain important information about them. We are particularly interested in the exact determination of the mean density of planets. So far, this has only been possible for a few hundred planets – and none of them resembles the Earth-Sun system. 

What are you particularly interested in?
Our big goal is to find planets that are habitable, i.e. that have conditions under which life could develop. Since we don’t really know how life develops, we have a lot of factors in mind. We cannot directly observe the biosignatures we are looking for, i.e. signs of life – exoplanets are much too far away from us for that. So we look for indirect traces. Our life on Earth today depends on a high content of oxygen in the atmosphere and on water. Therefore, we are looking for planets with surfaces where there is permanently liquid water and correspondingly moderate temperatures, as well as an atmosphere that is not too dense and not too thin. 

Are these the main criteria for Earth-like planets?
Yes. A central star that is similar to the Sun is also one of these criteria. However, we do not want to exclude the possibility of life in other constellations. However, with increasing distance from Earth as the only example known to us, it becomes increasingly difficult to provide circumstantial evidence for the existence of life. That is why we are initially focusing on Earth-like planets. In parallel, however, we are continuing to search for exoplanets with a wide range of properties in order to understand which types of planets exist at all and to investigate their habitability in the next step.

© OHB-System-AG

The PLATO space telescope (here an artist’s impression) is scheduled to launch from the Kourou spaceport at the end of 2026.

Do you already have first impressions?
Among the known five thousand extrasolar planets, there are types of planets that do not occur in our solar system. In general, the diversity of planets is far greater than we have long assumed. This raises new questions about their formation and habitability. Unfortunately, our instruments are not yet sufficient to detect an Earth-like planet around a star like the Sun. But this would be an ideal candidate for the search for life. So we cannot yet directly compare our planetary system with other systems. A first step towards changing this is the PLATO satellite mission. We can then study the atmospheres of the exoplanets found with PLATO with large telescopes such as the James Webb Space Telescope and its successor projects. 

More than eight hundred scientists from all over the world are working together in PLATO. How does the collaboration work?
Several consortia, each of which may consist of many partners, are in close exchange with each other. The overall PLATO mission is led by ESA, which also provides the launch vehicle, the ground segment of the satellite and contributions to the payload. The satellite bus, which carries the instrument, is being manufactured on behalf of ESA by an international industrial consortium. The international payload consortium of scientific institutes, also together with the space industry, is building the largest part of the instrument consisting of 26 cameras with associated electronics, on-board computers and power supply units. The payload consortium provides the data centre for the scientific processing of the data and organises the ground-based follow-up observations on telescopes, which will determine the majority of the discovered planets using the so-called radial velocity method. So it is important for the success of such a large-scale project that the various activities and the consortia and organisations involved mesh well. 

When do you expect the first results?
PLATO is scheduled to launch in mid-December 2026. This will be followed by a phase in which the functionality will be tested. Immediately afterwards, the observation of the first target field will begin. If everything goes well, we can expect the first data sets at the end of 2027, beginning of 2028. They will make it possible to characterise short-period planets. However, more time is needed to discover planets with long orbital periods. 

Will you still be leading the project then?
Missions like the PLATO mission are very long-term projects. PLATO was first proposed in 2009 and is based on ideas from even earlier projects. With projects like this, you have to think across generations. I myself will retire around the start of the mission and am looking forward to taking the project to the first data collection. Already, one of my tasks, together with the colleagues who launched PLATO, is to introduce the next generation of young scientists to this and subsequent missions.

© ESA

One of 26 cameras of the PLATO mission.

How did you get into this field of research?
I used to study comets, which can tell us a lot about the formation of our solar system. Then, when the first extrasolar planets were discovered in the mid-1990s, I switched to this line of research. Now, for the first time, we can compare our solar system directly with other systems and learn a lot about the processes that have influenced our system. And of course, I also find it fascinating to search for life beyond Earth. 

At the Leipzig anniversary conference of the GDNÄ, you captivated the audience with your lecture on extrasolar planets. What do such appearances mean to you?
When I give public lectures, I always notice how interested the audience is in our work. They want to know how planets are formed, how life is formed and whether there are planets with life around other stars. Today, for the first time, we can find answers to these questions using scientific methods – and I am extremely happy to report on this work to the interested public. 

In Leipzig, you got to know the GDNÄ’s student programme. Your research centre, the DLR, runs student laboratories, in which you are also involved. What is it about working with young people?
I think it’s important to show young people what research is really about, and I want to encourage them to think further. In my experience, astronomy is particularly suitable for this, because it deals with the big questions of where from and where to, which appeals to young people in particular. We can often motivate them to tackle even difficult courses in the natural sciences and engineering and to persevere until they graduate.