SDU must take a curious, bold and critical approach to generative AI

Since ChatGPT was launched in November 2022, it has been beyond doubt that this was the first step towards a technological paradigm shift. But can generative AI be disseminated at the University so that it will benefit all of SDU and be a help in everyday activities at the University?

These issues are being addressed throughout the organisation in parallel with the continuous evolution and use of the technologies.

We have asked University Director Thomas Buchvald Vind how he and the Rectorate view the potentials and the challenges of the deployment of generative AI at SDU. Here are his answers.

1. What is the Rectorate’s opinion on the use of generative AI at SDU?

Generative AI (GAI) has the potential to create major changes in society and at universities, and the technology already has a significant impact now.

Many different types of GAI have become widespread over the past 1½ years, and many different activities that make use of GAI are already taking place at the University. These will help mature our approach in both research, teaching and administration.

At SDU, we wish to create value for society and shape the future. As a university, we can create value by researching and teaching GAI and in that way help increase knowledge and skills in society. When we research and teach GAI, the University itself is also shaped by GAI – in both the scientific and the technical-administrative environments.

The Rectorate’s opinion on GAI is founded on the classical academic spirit.

• On the one hand, we must be curious about the new technologies and how they can create value for society and the University. We must be bold in checking and testing GAI.
  
  
• On the other hand, we must also be critical and aware of the maturity of the technologies and the pitfalls which may be challenging.

The use of GAI requires high technology comprehension among staff and students, so that responses and products from GAI are not accepted at face value.

Language models and AI are improving all the time, but responses may be biased. There may also be challenges related to copyright, information security and data protection.

Therefore, the Rectorate believes that GAI holds great potentials and that the University must take a curious, bold and critical approach to GAI.

2. Where does the Rectorate see the greatest innovative potential for the University?

The potential of GAI is great in both research, education and administration.

A few examples are:

• Improved efficiency – GAI can automate many administrative tasks, free up time for employees and improve overall efficiency. This efficiency is not only in relation to administration, but also to research (such as research applications and transcription) and education (such as feedback on teaching design).
  
  
• Learning/Education – GAI can be used to create personalised learning experiences, to adapt to individual needs and to provide more feedback to students.

• Research – GAI can be used to analyse data, generate hypotheses and design experiments, which may lead to new discoveries. In addition, all of this can be done on a completely different scale than what we have been used to.

If we are to exploit the potential, it is important that the technology does not stand alone. We must include the people (students and employees). We can create the best solutions through the combination of GAI and human insight.

3. A take on how generative AI will transform the University in the next few years?

GAI is a technology that is still maturing, and we will continuously address how to navigate opportunities and challenges.

We expect to see some clear and evident changes in the coming years in the administrative area, with more and more mature products entering our administrative systems landscape. This is especially true in relation to Microsoft 365 (Teams, Office, etc.), but we will also encounter and be able to provide completely new services.

We are also setting up a number of local GAI solutions that train on administrative SDU data. Over the next six months, we expect that this will provide us with relevant experiences to build on.

It may streamline anything from administrative case management and the way we communicate to the way information is searched for on mitSDU and SDUnet and much more...

In the field of education, in some areas it is already impacting the way we teach and conduct exams. Regardless of whether we embrace technology at SDU, the students do so.

GAI is already used for research purposes as well as for research dissemination and research applications/funding. The use of GAI will increase in the coming years, so that even more people will be exploiting the potential of the technology.

4. What are the biggest challenges from your perspective?

The main challenges relate to themes such as security, technology comprehension, human reaction, ethics and finances:

• Security – there are currently very few solutions that meet the usual requirements of the data protection framework. To be able to use GAI at SDU, it is important that you as a user familiarise yourself with the framework – see SDUnet
  
  
• Technology comprehension – as a user, you need to expand your digital skills in order to use GAI safely. This requires an understanding of GAI’s possibilities, but certainly also of the pitfalls and the maturity of the technology.
  
  
• Human reaction – quite understandably, there may be opposition to the use of GAI from some staff and students. For example, this may be due to concerns about job losses, concerns about privacy, concerns about the quality of the AI-generated output, or ethical considerations about the power of the technology.
  
  
• Ethics and philosophy – ethical considerations about, for example, bias and discrimination, but also philosophical reflections on what it does to people and society that we leave substantial parts of the creative world in the hands of technology.

• Finances – free solutions are typically paid for with access to your data. The solutions that are to be embedded in our system landscape do not come for free. The financial situation remains unclear. We have a special focus on uncovering this when we set up GAI solutions in Azure, etc. If all employees are to have Copilot in Microsoft 365, it will currently cost several millions per year in additional licenses. For that reason, we are actively concerned with where and with what solutions we can create sufficient value.

5. How can these challenges be met?

It is important to be very aware of the challenges and not just to be blinded by all the possibilities. Here are some perspectives on how we at SDU work with the challenges:

• Safety – it is vitally important that employees and students know the conditions for using GAI and stay updated on SDU’s guidelines in this area.

• Technology comprehension – courses are available, for example through the project AI in Education, and technology comprehension is a part of the courses that originate from AI in Administration, which have many participants. In addition, in the past year the focus has generally been on the work with digital skills (the competence wheel - in Danish only).

• Human response – involvement (for example, listening to concerns and resistance) and education contribute to critical and constructive work with generative AI.

• Ethics and philosophy – technology comprehension and dialogue are important.

• Finances – Trial actions and smaller initiatives that uncover the financial situation are a priority before we scale our initiatives.

6. How can we both elevate the organisation’s capabilities in generative AI and make digital security a high priority?

In addition to the above points, two other current projects in the digitalisation portfolio focus on securing the basis for using GAI to a greater extent. One is a project on rights management and the other a project on shared drives.

We must move forward at a speed that ensures that we do not compromise digital security. This places great demands on both the organisation and the individual employee, and it is a difficult balancing act, as many wish to make use of the possibilities of the technology. However, we will have to do it in a controlled way and within a certain framework