Conflicting Views on ChatGPT

기사승인 [401호] 2023.04.05  


   The advent of the latest artificial intelligence programming, known as ChatGPT-3.5 has caused a stir worldwide. The name evolved from combining the words 'Chat' and 'Generated Pre-trained Transformer.' As the origin of the name implies, it is an artificial intelligence chatbot specializing in written text where users communicate with the program through a text box and receive written replies. It offers up common sense responses instantly. ChatGPT has been so successful that two months after its initial release, it had amassed 100 million monthly users. This is the fastest recorded growth of a new program as compared to other social media such as TikTok or Instagram. According to the Union Bank of Switzerland, the software had the highest rate of daily users recorded with 13 million subscribers. Amid the constant demand for ChatGPT, OpenAI announced a pilot subscription plan for ChatGPT Plus, a premium version of the freeware.

▲ Official Website of OpenAI's ChatGPT (Photo from OpenAI Website)

   ChatGPT can easily be understood as an AI text generator that can answer all types of questions and complete the official website, it can even answer follow-up questions and accept feedback on its previous responses. Conversely, if there is any misinformation from the user, it makes the necessary corrections. In addition, it uses an AI-based moderation system called ‘Moderation API,’ which can reject discriminatory, aggressive, or inappropriate questions. On March 14, OpenAI, released a newer version of its Large Language Model (LLM) ChatGPT-3.5 called GPT-4. With this upgrade, the AI program can also accept images as input and generate captions, classifications, and analyses. Most users are saying that ChatGPT-4 is exceptional enough to replace some professions. It creates literary work such as poems, novels, and even plays. As a result, many enterprises are already actively using it. Johannes, a real estate broker in Iowa, entrusted the software to author its property listings, and was happy when it completed the task in five seconds. If he had done the work himself, it would have taken much longer. U.S. real estate firms are using ChatGPT to promote their listings on SNS, welcome new customers, and even calculate mortgage loans. CNN reported that it will be difficult to return to our former ways of operating before AI landed in our laps. Korean domestic coffee chain Ediya Coffee used ChatGPT in their recent advertising campaign. Ediya Coffee posted a dialogue between ChatGPT and a marketer on its official blog. The marketer asked it about the company and then asked it to write an advertisement for their new menu. Surprisingly, it delivered an almost perfect answer, producing advertising phrases appropriate for the marketer. While positive reviews of ChatGPT are pouring in, ChatGPT Plus, the paid subscription program, is also in the limelight. OpenAI said, "Its users have prioritized access when the web traffic is high. In addition, it will answer faster." OpenAI expects that this will also open-up more space for users who rely on the free version.

   Although ChatGPT is receiving considerable attention and has already been adopted by numerous industries, there are persistent concerns that it still an AI chatbot. First, there are functional problems. Given that the program’s primary language is English, many grammatical and contextual errors are transmitted when used in languages other than English. In addition, many users complained responses were unfiltered and ambiguous texts, making them unreliable for use. This is because it searches, learns, and provides information that indiscreetly exists on the Internet, instead of focusing solely on specialized information by experts. The second problem being raised is in the field of education. With the help of ChatGPT, students in their teens and 20s who write reports will not get an accurate evaluation of their knowledge. If they prompt the program for the information they need to write, and insert a few keywords, it will instantly create a completed essay, making it difficult for educators to evaluate how much a student has actually learned. In fact, this fear is so great that the New York City Department of Education banned the use of ChatGPT in all public high schools in early January, by preventing access to it from school Wi-Fi networks. This move was criticized as unfair because the ban applied to free access sources relied upon mostly by underprivileged students. This would mean those who could afford it, could still access it elsewhere. There is also the problem of unintended information leaks. Companies, including Walmart and Amazon, released policies that prohibited employees from using it to post their own programming code or generate new codes, insert proprietary or confidential information, and more.

▲ There are also concerns about the misuse of ChatGPT. (Photo from Unsplash)

   To understand more about the learning system of ChatGPT and changes that can be expected due to future developments in AI, the Dankook Herald interviewed Professor Kim Dong-jae from the graduate school Engineering Division on the Jukjeon Campus, whose primary field of research is AI-based convergence. Unlike the existing open-source software models, ChatGPT is a specialized pretraining model, generating human-like words and sentences naturally. Professor Kim said one of its benefits is that it helps users increase productivity. When ChatGPT aids people in creating simple codes, it allows them to concentrate more on more important areas. Even though some significant source codes produced by the program need to be revised, simple ones are produced with relative ease and assist people by reducing their time working on problems. However, according to Professor Kim, LLM, like ChatGPT, has a hallucination effect, in that is has a tendency to fabricate facts or false information. Also, the more data that is created, the more learning is required. There are no specific regulations that protect the copyright of the information collected. Nevertheless, he believes that the proficient usage of AI would usher in a new era, where easy tasks are delegated to machines.

   With the advent of ChatGPT, the benefits to humans were immediately obvious, and its many appealing features led to widespread interest. Nevertheless, certain ethical issues may be neglected or exacerbated if their technology growth results in our complete dependency on artificial intelligence. For ChatGPT to be adopted safely, we must acknowledge its existence and guide its use in a positive direction without abusing it for inappropriate purposes.

김주연, 이진희, 김지원

<저작권자 © The Dankook Herald 무단전재 및 재배포금지>




1 2 3

섹션별 인기기사 및 최근기사