hankyoreh

(Getty Images Bank)

Poet Shin Dong-ho was recently astonished to see a poem posted by a fellow poet on social media. Titled “A Gray Afternoon in February,” it had been requested by the other poet from the artificial intelligence-based chatbot ChatGPT — and it was not bad.

What stood out in particular was a line that read, “The trees are bare, lacking self-respect, stretched out as though with hidden branches.”

“I don’t know if it was aware of something or not when it used the phrase about the trees ‘lacking self-respect,’ but it was quite convincingly written,” said Shin, 59, in a telephone interview with the Hankyoreh.

“I was astonished that an AI program had used the word ‘self-respect,’” he added.

He also predicted that “even more astonishing poems are going to be written as [the program] acquires more and more references.”

“For a poet whose job involves working with language, I felt threatened,” he said.

In the two months since being unveiled by the startup OpenAI on Dec. 1 of last year, the conversational AI chatbot ChatGPT has been drawing major attention, with over 100 million users around the world.

Users ask “questions,” and ChatGPT provides the “answers,” assembling vast volumes of information to create sophisticated and logical texts.

But as ChatGPT proves itself capable of writing academic papers and even creative works such as poetry, many are voicing concerns that it could take the place of human professions.

“Technology is never going to completely replace [people] in something like psychiatric medicine, which requires empathetic capabilities,” said a 41-year-old psychiatrist surnamed Jang, adding that there “seems to be ample potential for advancements in medicine.”

“I asked ChatGPT whether one of the tranquilizers commonly prescribed in psychiatry is dependency-forming, and it replied, ‘It may cause dependency, and if long-term usage is halted, an acute response may occur,’” Jang recalled.

“I agreed, and I also felt nervous [as a physician],” they added.

Legal services are facing a similar threat.

Kang Tae-hyeon, head of the legal AI team for the law firm Hankyul, said, “While it’s unlikely to take over completely any time soon with duties like writing documents that involve complex issues, it might be able to substitute once issues have been resolved with the time needed for development and recognition of the AI’s liability.”

“It is possible to make simple inferences that ordinary people want in first-order logic, such as [answers related to] a response to a hit-and-run accident,” commented Rhim Young-yik, CEO and lawyer at Intellicon Meta Lab, a law firm.

"However, it seems that complex inferences made through second-order logic for individual cases are difficult to do directly and that only [answers providing an overall] review are possible. It takes a lot of time and money to make second-order logic inferences directly and issues such as social ethics and systems are also interconnected,” Rhim explained.

With AI like ChatGPT becoming a part of our everyday life, changes at schools where tests take place will also be inevitable.

"Its Korean is still at a low level, but its English is at an almost perfect level. When writing a paper, there are cases when you need to pay an outside service for it to be translated or for the [English] draft to be checked, but if you use ChatGPT it seems such costs will be significantly reduced,” says Shin Seung-yoon, 21, a student at Ulsan National Institute of Science & Technology.

Some schools in the US block access to Wi-Fi networks and the use of ChatGPT on campus while some universities are also prohibiting the use of computers when submitting exams and assignments and are instead going back to handwritten and oral tests.

But many say that ChatGPT is not yet at the level where it should warrant concern of possibly replacing humans in certain professions.

“The current impact of ChatGPT is similar to the feeling we had when we first encountered the Internet,” says Hong Ki-hoon, a professor of business administration at Hongik University.

“It seems like all the information being given is new, but in fact it’s just providing existing information. It’s just that this information couldn’t be found until now,” Hong said. “It is clear that the paradigm of how to find information has changed, but it is not mature enough to replace humans.”

Similarly, Kim Seung-joo, a professor at Korea University's School of Cyber Security, shared, “ChatGPT is not that different from Naver’s jisik-iN [community-driven Q&A platform] or Wikipedia. You’ll find countless reports that are written using information found there, but just as we know that not all those answers are correct, we also shouldn’t see all of ChatGPT’s answers as correct.”

There has also been positive feedback regarding ChatGPT’s use for studying English. Since ChatGPT’s original operating language is English, it can provide immediate help with practicing the language.

In particular, considering how many people have to spend money to learn how to write natural-sounding English sentences or to have their writing corrected, a tool like ChatGPT is of great help for improving one’s English writing skills.

According to a 44-year-old office worker surnamed Kim, while existing English writing apps use slightly unnatural expressions, “ChatGPT writes natural English sentences that are close to perfection.” Kim also explains that it’s possible and helpful to ask ChatGPT for more than one version of the same expression.

For example, if you instruct it with specific commands such as “have a conversation with me and correct me when I use a wrong expression,” you can have a perfect one-on-one conversation and also learn new conversational expressions.

Those who first discovered this way of using ChatGPT on Youtube left comments “I only used it for coding but never thought of using it in this way.”

By Kwak Jin-san, staff reporter; Kang Jae-gu, staff reporter

Please direct questions or comments to [english@hani.co.kr]