This blogpost was written as an opinion piece by our managing partner, Søren Boesen.
A few weeks ago, the chatbot ChatGPT was launched. It is unlike anything seen before as it can understand and create text at an impressive level. It is capable of generating text on virtually any topic, in any style. Summaries, physics reports, analyses, Danish styles, poems, even letters to readers. From just a few words of input. I know this because I have used ChatGPT intensively since it was released and have been using the underlying language model GPT3 for the last six months.
It's insanely fascinating and wildly terrifying.
The technology feels magical and like something from the distant future. But it's here right now and anyone can access it. Soon it will be embedded in our writing programs and be ubiquitous.
The technology has great potential for education. Imagine having a teacher who has read all the books in the world and who is both super intelligent and always available to help. All school students have that now. But every technology comes with its dilemmas and challenges. For ChatGPT and similar language models, they are obvious and serious: Cheating in assignments and in exams.
I'm sure students at all levels of education are already using the technology. And with the latest update, it's now even more powerful. I know this, because this blogpost was partly written by chatGPT.
Cheating is nothing new, but what is new is that this kind of cheating is virtually impossible to detect. ChatGPT has no specific "signature". It is trained on a large corpus of text and it generates unique results. Additionally, it can be customised and fine-tuned to generate text specific to a given task, making it even more difficult to determine whether it was used to generate a style or task.
It therefore gives students the opportunity to pass tests and get good grades without actually learning and understanding the syllabus. This will have serious consequences for the quality of education, for the students' learning and the value of grades.
If this is not to happen, action must be taken now.
Teachers must know about the technology so that they can be aware of whether it is being used, enter into a dialogue with the students and adapt the teaching. Tests and exams should be adjusted and ensure that grading is based on oral tests and written tests without internet access.
If we don't act now, artificial intelligence can easily lead to human stupidity.
GPT3 was used to create this blogpost, which was then edited by the human co-author.