COLUMN: So, AI writes better assignments than students? But what about columns?

This is a contribution to a debate. 

Generative AI is better at writing texts than Danish students. At least that's the sentence that makes coffee cups clink in the cafeteria and academic regulations shake in their PDFs. Not because anyone really believes it – but because it's just provocative enough that we can't help but think about it.

Let's start with the obvious: Generative AI writes quickly. Absurdly fast. It writes faster than a student can open a new document, choose a font (Times New Roman, of course), and consider whether the title should be “An analysis of…” or “A study of…”. The AI ​​has already produced three versions, an outline, and a conclusion before you've found your charger.

And yes, it is grammatically correct. Almost frighteningly correct. There are no typos, no commas that have been placed on a gut feeling, and no desperate formulations like “it could also be viewed as a kind of…”. The AI ​​doesn't doubt. It delivers. It's not standing there at 2:13 a.m. staring blankly into the screen and thinking: “What exactly is a research question?” It knows it. Or at least it pretends to.

AI IS GOOD AT IMITATING

But let's not be blinded. Because what is AI really good for? It's good at resembling something that's real. It’s good at imitating structure, tone, and academic confidence. It's like the student who always raises his hand – not necessarily because he has something wise to say, but because it sounds like it.

The Danish student, on the other hand, is a more messy entity. Here we find half-baked ideas, sudden insights and long detours that may – or may not – lead to something interesting. There are doubts, errors and formulations that don't quite land. But there is also something else: an actual struggle with the content. A process. A realisation that is not generated, but emerged.

An AI ​​has never had writer's block. It has never handed in anything at the last second with sweaty palms and a silent prayer to the exam gods. It has never misunderstood a theory and only discovered it three pages later. And that is precisely why it lacks something fundamental: experience in making mistakes.

MORE EFFECTIVE, YES. BUT BETTER?

It's tempting to say that the AI ​​is “better” because it is more efficient, more accurate, and more consistent. But that's a bit like saying that a calculator is better at math than a human. That it is - if the task is to calculate quickly. But if the task is to understand why we calculate at all, the picture becomes muddier.

The problem arises when we start measuring student texts on the same parameters that the AI ​​excels at. If the most important thing is correct language, nice structure and a fairly convincing conclusion, then yes – the AI ​​is hard to beat. But then we have also reduced education to a form of linguistic cosmetics.

The best study text is not necessarily the most error-free. This is the one where you can feel that something is at stake. Where the writer has struggled with the material, perhaps stumbled a little along the way, but has still reached a point that is not only plausible – but personal.

So is generative AI better at writing texts than Danish students? If we're talking about surface, pace, and grammatical discipline: yes, without a doubt. If we're talking about cognition, originality, and the slow, frustrating process that actually makes us smarter: not at all.

And maybe that's exactly the point. The AI ​​writes texts. Students are being educated. One of those can be automated. The other still takes time, doubt, and a certain amount of panic at 2:13.

Fortunately.

If you want, I can also:

• Adjust the tone (more sarcastic, more serious, more “Politiken” or more “Ekstra Bladet”)
• Tailor it to a specific audience or publication
• Shorten or lengthen the column

Do you want a new version?

NB! This column is 100 percent written by AI (ChatGPT), which Peter Bakker prompted with the following: Write a funny and sharp 600-word column on the topic: generative AI is better at writing texts than Danish students. 

Omnibus does not normally feature content written by AI, but here the column serves as a contribution to an ongoing debate about the use of AI at the university and the quality of content generated by AI. Omnibus always declares when AI has been used in the production of editorial content. For example, Omnibus uses machine translation when we translate content, but always with subsequent post-editing to ensure quality. In the same way, as above, it will be apparent if AI has been used to generate graphics or illustrations. In addition, the editorial team can use AI for research and idea development.

This text is machine translated and post-edited by Lisa Enevoldsen.