고등학교 공통영어2 (미래엔 김성연) 연계 글

Last month, my history teacher asked us to write a short report on the old markets of our city. I asked an AI chatbot to suggest three books on the topic. Within seconds, it gave me three titles, each with an author, a publisher, and a year. At first, I found it amazing that a program could do an hour of library work so quickly. It even offered a quotation from the third book, and I copied it into my first paragraph.

The next day, I searched the online catalog of my school library. The first two books were there, but the catalog had no record of the third one. The city library's catalog did not list it either, so I asked our school librarian for help. In the end, she concluded that the book, "Lanterns of the Salt Market" by Harold Wenbrook, did not exist. The chatbot had made up the title, the author, and even the quotation. The fact that the book came with a publisher and a year had kept me from doubting it.

Why would a chatbot invent a book? A program of this kind builds its answers by predicting which words are likely to come next. Predicting likely words is different from checking facts, so a false sentence can sound perfectly natural. Researchers call this kind of believable but false output "hallucination." Librarians at several universities warn that chatbots may list sources that do not exist.

I now consider it my job to check every source before I use it. First, I search a library catalog to make sure the book or article exists. Next, I find the original text and read the part I want to quote. Finally, I compare the main facts with another trusted source. These steps take time, but they have made it easier to spot answers that only sound right.

In my view, a chatbot is still a useful tool for starting homework. However, I no longer accept the idea that a confident answer must be a correct one. I also keep in mind the possibility that even a detailed answer may be false. I think it wise to treat every chatbot answer as a first draft.