By Karsten Kryger Hansen, Senior Consultant Aalborg University Library.

I will focus on the use in administrative practice. Research and studies have other scenarios and usage potential as well as some other issues in terms of good practice, which require an article of its own.

Not a search engine, but also a kind of search engine!

Large language models are a completely new way of accessing knowledge and information. The great strength of Copilot and corresponding platforms is that they are ‘generative’. They allow - with all the uncertainty it gives – to create new content. When you formulate a ‘prompt’ you describe the task at hand, the target audience, what you want to achieve, writing style etc. Afterwards, you have a ‘dialogue’ with Copilot about adapting, perfecting, and changing the result. In that way, Copilot is a sparring partner for solving an assignment, much more than it is a search engine – although it also uses searches.

Due to the language model being creative, it can ‘hallucinate’. This is a feature-by-design, which makes it strong, but means it must be handled with appropriate understanding. One should be careful in securing that Copilot's answers are true and adequate.

Break habits

For most of us, the way we solve assignments is just routine. Thats good because it means we don’t have to constantly think carefully about how to solve tasks. However, it also means that copilot and other tools can have difficulties entering these routines. At least while it is not deeply integrated into the tools – which we will later see with Copilot 365.

That’s why, for a transitional period, it can be good to force oneself to try Copilot, before starting an assignment. Use a bit of extra time – albeit it is a sparse resource – to explore whether Copilot can contribute to solving the task. Maybe it can assist with formulating an email, understanding a term, analysing a piece of text, breaking down an assignment, translating, making a draft for a seminar program, giving feedback on nuances for a written text, or other things.

Get a good start with Copilot

There are three ways to acces Copilot:

You can read more about the various ways to access Copilot on the University Library’s web pages about generative AI.

It is easy to start. Here are the first steps. Try Copilot next time you have to…

• Find an illustration for a PowerPoint, by requesting Copilot to generate a picture to illustrate a point.
• Plan a theme day about a specific topic, and search for perspectives and methods to work on the actual day.
• Make functions, macros or other things in Excel or other programming languages.
• Make a text and use Copilot to receive feedback on a text you have written, where you also give as input who the target audience is, the intention and more.
• Understand a term or a text, then you can request Copilot for an explanation.
• Brainstorm content for a new webpage, newsletter etc.

Guidelines for the use of AI

A workgroup at AAU has written a code of conduct for the use of generative AI in administration. This code doesn’t explain exactly what you can or must do, but sets the framework for what you should consider before jumping into Copilot and generative AI as a whole in terms of responsibility, data protection etc.
Read it here – and remember, it must be read before using AI for administrative purposes (in Danish):  AI kodeks for administrativ opgaveløsning