AI bot is a threat to education, though

UU students: ‘ChatGPT is not going to take over the world’

Chat GPT Foto: Shutterstock
Photo: Shutterstock

DUB took a walk in the Utrecht Science Park during the latest exam week, wondering if UU students will still study for their exams now that ChatGPT took the world by storm.

We ran into Bram, a student in Educational Sciences, who says ChatGPT was a big hype a few weeks ago but its popularity has waned somewhat ever since. He adds that, personally, he isn’t a fan of it, and then rushes off to his exam – for which he studied hard, by the way.

Even so, students are definitely making the most of the new technology. Mateo, a Master’s student in Bio Inspired Innovation, turned to the chatbot to get a quick answer whenever he got stuck in the course Introduction to Python. The tool saved him a lot of time.

ChatGPT uses prompts – short textual input – to generate pieces of code. “That means you don’t have to spend hours searching for the right information online. I often get exactly what I need or at least something in the right direction. Maybe that’s because this course mainly focuses on basic techniques.”

Mateo already knew by early January that the course would have an open-book exam, in which students would have access to the Internet — which means students they would have access to ChatGPT. The rules of the course couldn’t be adjusted at such short notice.

Quizzing yourself
UU's Executive board was as startled as anyone else by the speed with which AI has been advancing and rushed to tighten its plagiarism rules. Nevertheless, the goal of the new regulations isn’t to keep ChatGPT from ever being used but rather to regulate its use. Teachers will mainly have to change the way their courses are assessed. For now, things haven't gone that far.

Although Mateo was still allowed to use ChatGPT for his "old-school" exam, he studied hard for it anyway. “I might use ChatGPT as a last resort,” he swears. “It still makes a lot of mistakes. Besides, it’s been unavailable quite often lately because the server is overwhelmed. So, if you didn't study, you won’t make it.”

Mateo seems to be using the new developments responsibly. But that doesn't mean there are no students using the software to write entire essays and assignments.

“A friend of mine works as a French teacher and he says that around 50 percent of all papers are written by AI now,” we hear from Caitrione, a History student. “I know people who have used it to write essays. The biggest danger is that people will become very lazy and use ChatGPT instead of thinking for themselves. On the other hand, this technology has its benefits as well. I’ve heard that there are people using it to quiz themselves ahead of an exam, for instance.”

Hard to keep up
For now, papers and essays written by ChatGPT are rather easy to recognise because of certain sentence structures and abrupt changes in language. Students still have to put in some work to smooth out these issues. But AI has been improving fast and the expectation is that ChatGPT's output is going to be harder and harder to distinguish from authentic work.

Johan Jeuring, a Professor in Information and Computing Sciences, acknowledges that ChatGPT will bring forth considerable challenges for the education sector. “Assessment methods like traditional essays and open-book exams are less and less useful this way,” he states.

“Besides, the technology is advancing so fast that the university is having a hard time keeping up. You can’t just adjust the exam rules halfway through a term. That's why I tend to believe that many students are using it to make the work a little easier for themselves and get higher grades.”

However, Jeuring agrees with the university’s principle of not engaging in an "arms race" between artificial intelligence and anti-plagiarism tools. “We shouldn’t focus on setting up stricter plagiarism controls but rather on how we design our assessments.”

Taking over boring tasks
Jeuring thinks that the new technology could also have positive effects, as long as students learn how to use it in the right way. “I recently made ChatGPT take one of my programming exams, and it failed. But after I made a few adjustments, it got a passing grade. I was also done much faster with my work. So, the programme can be a supporting tool, taking over easy, repetitive tasks, but you do have to be capable of working with it. By means of comparison, a calculator is useful but only if you understand what you’re doing with it.”

Jeuring observes that students need to master the material before they’re allowed to use ChatGPT. “It isn’t hard to limit access to the Internet during exams. We’ve got systems in place to do that. So, if you let ChatGPT do your assignments for you, you’re only hurting yourself. But if students have the necessary knowledge and skill, ChatGPT can come in handy.”

He thinks it’s important to integrate ChatGPT into the courses so that students learn to use the technology appropriately. “For example, you could give them assignments that make them use ChatGPT, and then evaluate that input and improve it. After all, its answers are far from perfect.”

Most students are well aware that chatbots aren't perfect. And they say that’s not the only downside. Unlike traditional information providers, ChatGPT doesn't reveal its sources.

“When you ask for its sources, Chat GPT says it cannot give them because there are too many of them,” explains IT student Eke. “The bibliography on Wikipedia was already pretty dubious but this is much worse. You basically have to verify everything it says by Googling it and remain extremely critical.”

Concerns about the future
What do UU students think about the potentially destructive effects of artificial intelligence in other fields? Google sees the programme as a threat to its own existence, while Elon Musk believes that AI programmes like ChatGPT might become smarter than humans and more dangerous than nuclear bombs.

Most students we interviewed at the Utrecht Science Park aren’t afraid that the chatbot will develop a consciousness of its own and take over the world someday. To Merlijn, a Master’s student in Computing Science with a Bachelor’s in Artificial Intelligence, the question of whether computers will ever be able to have their own consciousness is a philosophical discussion. Hence his decision to follow a Bachelor's that included Philosophy courses.

“AI without its own consciousness is called ‘Weak AI’. If it does have its own consciousness, it is called Strong AI. People who believe in ‘Strong AI’ are often convinced that consciousness can only be a consequence of biological processes in the body. From that point of view, you could create a consciousness by mimicking those processes in a computer or else. But I believe there’s more to our consciousness than just biological processes. I’m not sure what it is exactly but I don’t believe it can be recreated with a computer.”

Merlijn is thus convinced that there is no such thing as ‘strong AI’ right now. Moreover, he says most IT students don’t believe such a thing would be possible in the future. “It’s not a realistic to think that a chatbot like that will take over the world.”

For now, students do have other concerns, such as whether people will stop being critical enough, believing the chatbot’s output to be the truth. “People already believe a lot of what’s posted on social media,” Merlijn ponders. “That’s because they don’t understand how it works. They just need good information and education about these new, complex techniques.”

Other students are afraid that ChatGPT will replace a significant number of jobs down the line, make their studies redundant. Professor Jeuring, however, debunks that scenario. “People said the same in the early 1990s about IT development but that ended up creating more jobs instead. This type of tool can only take over relatively easy, repetitive tasks. I believe this new technology will enable us to focus on creative, more interesting tasks.”

After all, this type of chatbot cannot be creative by nature, stresses the professor. “They only generate output based on previously-existing information. Perhaps the output can pass plagiarism checks but these technologies basically never create anything new.”

Nevertheless, Naomi, an Audiovisual Media student at HKU, does worry about the creative industry, where she and other students would like to work. ChatGPT can already write simple stories and song lyrics; other AI tools make music and art. “I just really hope people will keep appreciating it when something is created by a human,” she sighs. “When something is human-made, there is a certain feeling behind it. Someone’s thought about every single stroke. That doesn't happen with AI art.”