-
- Returning Students Lament the Rise of ChatGPT Note-Taking… A Campus Transformed by AI NEW
- [Returning Students Lament the Rise of ChatGPT Note-Taking… A Campus Transformed by AI] With the spread of generative AI tools like ChatGPT, university students are actively using AI for a variety of academic tasks—including writing reports, summarizing lectures, and conducting research—significantly changing the landscape of learning on campus. Students are sharing methods among themselves for capturing lecture materials and using ChatGPT to summarize them or quickly complete assignments with the help of GPT. Professors are gradually adapting to this trend as well—for example, Professor B from the Department of Journalism and Mass Communication at a national university has even permitted unrestricted use of AI in class. In the professional world, there is a growing preference for students who are proficient in using AI. Since the quality of assignments can differ even with the same tool, the ability to use ChatGPT effectively is increasingly regarded as a competitive edge. Source: "챗GPT 필기 등장에 복학생들 푸념…AI로 바뀐 캠퍼스", 문화일보, 2025.04.20.
-
- 작성일 2025-04-24
- 조회수 8
-
- AI Detection Tools Are Not the Answer to Cheating in Classrooms – A New Way of Assessment Is Needed
- [AI Detection Tools Are Not the Answer to Cheating in Classrooms – A New Way of Assessment Is Needed] With the rapid development of AI (Artificial Intelligence), schools are also trying to use AI in many ways. One common method is using AI detection software to stop cheating. However, recent research and teachers say this technology is not always reliable or effective. AI detection tools usually analyze something called “perplexity” and “burstiness” to decide if a sentence was written by AI. But this method is not perfect. It can be unfair to students who are still learning English. Sometimes, the tool can even say a human wrote text with AI, which is called a “false positive.” For example, Turnitin’s AI detector says it is 99% accurate. But even a 1% false positive rate can be a big problem. At the University of Winnipeg in Canada, students submit around 250,000 assignments every year. If 1% are marked wrongly, that’s about 2,500 false cheating cases. According to a survey by the Center for Democracy and Technology, 68% of public school teachers in the U.S. (K–12 level) said they use AI detection tools. Some schools in Canada are also doing the same. But relying too much on AI detection can cause stress for both students and teachers. It also does not solve deeper problems with how schools test and grade students. Professor Michael Holden from the University of Winnipeg said that many assignments are now done at home, without a teacher watching. In this case, it's hard to know if the student really did the work themselves. He said we need to change how we assess students, instead of just making AI detectors stronger. Experts suggest using a type of grading called "formative assessment," which looks at the whole learning process — like drafts, progress, and feedback. They also say it’s important for students to take part in class and work on real projects where they solve problems by themselves. Professor Holden said, "In the age of AI, education needs to change how we evaluate students. We must understand what students truly know and what skills they really have. That means moving to new ways of assessment that focus more on the learning process and students' abilities." Source. : "[인공지능 줌인] AI 감지 소프트웨어, 교실 부정행위 해결책 아니다… 평가 방식 전환 필요", WIKI KOREA, 2025. 3. 6.
-
- 작성일 2025-04-17
- 조회수 26
-
- Dalseo District, Daegu, Accelerates ‘Smart Administration Innovation’ with ChatGPT Implementation
- [Dalseo District, Daegu, Accelerates ‘Smart Administration Innovation’ with ChatGPT Implementation] Dalseo-gu in Daegu is officially launching an AI-driven smart administration system using ChatGPT. This initiative aims to improve administrative efficiency and provide faster, more accurate public services for residents. To digitalize administrative tasks, the district is introducing several AI-powered chatbots. These include a chatbot that provides essential administrative information, another that automatically generates press releases and official speeches, and a translation chatbot to assist multicultural families and foreign residents. The district is also focusing on automating repetitive tasks to boost productivity. For example, it has successfully reduced the time required for monthly overtime meal expense calculations from 30 minutes to less than one minute through automation. Moving forward, Dalseo-gu plans to expand the use of ChatGPT-based automation programs to maximize administrative efficiency. To help employees develop AI-related skills, the district will conduct eight customized training sessions annually. This training will support employees in applying AI technology to their work, accelerating smart administration innovation. Lee Tae-hoon, the mayor of Dalseo-gu, stated, “With this AI adoption, we will provide fast and efficient administrative services and establish a leading model for smart administration.” Source. : "대구 달서구, ‘챗GPT’ 도입해 ‘스마트 행정 혁신’ 속도낸다", The Segye Times, 2025. 3. 2.
-
- 작성일 2025-03-14
- 조회수 105
-
- OpenAI Releases ‘GPT-4.5,’ Its Final Non-Reasoning AI Model
- [OpenAI Releases ‘GPT-4.5,’ Its Final Non-Reasoning AI Model] OpenAI unveiled its latest artificial intelligence (AI) model, GPT-4.5, on the February 27th (local time). This model is the successor to GPT-4o, which was released in May 2024, and is known to be the last non-reasoning model. OpenAI stated that GPT-4.5 has improved pattern recognition and association analysis capabilities compared to previous models. It also has enhanced ability to generate creative insights without logical reasoning. Additionally, with improved emotional intelligence (EQ), it enables more natural conversations and is expected to be highly useful for tasks such as writing, programming, and practical problem-solving. The model is also expected to reduce hallucinations—instances where AI provides incorrect or misleading information. Benchmark results show that GPT-4.5 has a hallucination rate of 37.1%, which is lower than its predecessor, GPT-4o (59.8%), and the latest reasoning model, o3-mini (80.3%). Moreover, its accuracy in answering simple questions stands at 62.5%, surpassing GPT-4o (38.6%) and o3-mini (15.0%). OpenAI CEO Sam Altman stated on social media that GPT-4.5 would be released “in a few weeks” and described it as “the last model we develop that does not follow the chain-of-thought approach.” The chain-of-thought approach refers to an AI's reasoning process, where it goes through intermediate steps to derive an answer. This suggests that GPT-4.5 will be the final model that operates without such reasoning. Source. : "오픈AI, 마지막 비(非)추론 AI 모델 ‘GPT-4.5’ 출시", IT Chosun, 2025. 3. 2.
-
- 작성일 2025-03-14
- 조회수 103
-
- Developers like ChatGPT more than any other tool
- [Developers like ChatGPT more than any other tool] In May 2024, StackOverflow conducted a survey of 65,000 developers worldwide to understand their preferences for development tools, experiences, and work environments. The results show a significant increase in the use of generative AI tools. When asked which AI search and development tools they had used regularly in the past year, 82% of respondents said they used ChatGPT, making it the most popular choice. Other commonly used tools included GitHub Copilot (41.2%), Google Gemini (23.9%), Bing AI (15.8%), and Visual Studio IntelliCode (13.6%). When asked which AI tools they planned to continue using in the future, 75% of developers chose ChatGPT, the highest among all options. Even among those who had not used it yet, 63.6% said they wanted to try it in the future. Experts believe ChatGPT’s popularity is due to its strong performance, even in its free version. The use of AI tools in development processes is also growing. 76% of developers said they are currently using AI tools or planning to use them, up from 70% last year. Additionally, 72% of respondents had a positive view of using AI tools in their development work. However, opinions on the accuracy of AI tools were mixed. When asked if they trusted AI-generated outputs, only 43% responded positively, while 30.4% had a negative opinion, and 26.6% were neutral. Regarding whether AI tools handle complex tasks well, only 3.3% strongly agreed, while 32.7% felt AI was helpful but not great for complex tasks, and 43.2% said AI tools performed poorly in such cases. Companies and teams also faced challenges when using AI tools. The biggest concern was a lack of trust in AI-generated outputs (66.2%), followed by insufficient understanding or trust in the codebase (63.3%). Additionally, 30.7% of respondents mentioned a lack of training as an issue. These results suggest that most developers do not see AI tool limitations as user errors but rather as technical problems that need improvement. Source. : "개발자는 챗GPT를 좋아한다, 그 어떤 도구보다", 바이라인네트워크, 2025. 1. 7.
-
- 작성일 2025-02-14
- 조회수 117
-
- Professors who used ChatGPT experienced a significant improvement in the quality of their English papers
- [Professors who used ChatGPT experienced a significant improvement in the quality of their English papers] Recently, many scientists have started using ChatGPT for tasks like writing research papers and searching literature. On December 24th, Nature reported the results of a survey conducted by the international academic publisher Wiley. Out of 1,043 respondents, 81% stated they use ChatGPT either personally or professionally. Additionally, Nature explained that last year, over 60,000 scientific papers directly benefited from AI assistance, accounting for just over 1% of the papers surveyed. The field of computer science showed the highest usage of AI, with 17.5% of papers created with AI assistance. In biology, over 10% of papers were estimated to have used AI during the writing process. Gabe Gomes, a chemist at Carnegie Mellon University in the U.S., said, “Large Language Models (LLMs) easily break down language barriers.” In fact, Nature highlighted that ChatGPT is especially helpful for individuals who use English as a second language. There is also a growing number of AI services designed to assist scientists with research and paper writing. Google’s AI agent, Deep Research, provides step-by-step research plans when asked about a specific topic. It also finds necessary information and organizes it into a report format. Japan’s AI startup, Sakana AI, offers a solution called AI Scientist, which uses LLMs to autonomously handle the entire research process, from generating ideas and conducting experiments to summarizing results, writing papers, and even reviewing them. In Korea, the Korea Atomic Energy Research Institute has introduced Atomic GPT, a specialized LLM tailored to the nuclear energy field. It has been trained on general knowledge, publicly available research papers, and various reports in the nuclear domain. Source : 챗GPT 활용한 교수들, 영어 논문 질 확 뛰었다, HANKYUNG, 25 December, 2024.
-
- 작성일 2025-01-03
- 조회수 203
-
- 5 Tips for Using ChatGPT Effectively, Shared by OpenAI's Product Lead
- [5 Tips for Using ChatGPT Effectively, Shared by OpenAI's Product Lead] On the 16th, Nick Turley, the product lead at OpenAI, appeared on the YouTube podcast "Hard Fork" and shared five ways to enhance productivity using ChatGPT. He mentioned that by thoroughly exploring the many features added to "GPT-4o", users can discover some highly useful tools. The first recommendation was using the "Advanced Voice Mode (AVM)." This feature enables natural conversations, unlike traditional text-based interactions. Users can adjust intonation to emphasize specific parts of their questions and even interrupt the chatbot mid-response without waiting for it to finish. He noted that this feature is available for free to users through the Windows PC app. The second tip was not to use ChatGPT solely for asking questions but also to share information you want it to know. ChatGPT includes a "memory" feature that allows it to remember chat details and reflect them in future conversations. Users can turn off memory or even instruct the chatbot to forget specific details. This feature was described as becoming more valuable the more it is used. Creating a custom chatbot through the "GPT Builder" was highlighted as another useful method. While most people might think this requires significant effort, he explained that it is surprisingly simple and takes only seconds. Additionally, users can customize everything they want and keep the bot private. The fourth tip was "file upload." He introduced this feature as incredibly useful when reading lengthy papers or manuals. It not only summarizes texts but also extracts specific information or quotes and compares content with other documents. With a context window size of 128,000 tokens, GPT-4o can process an entire book at once. Lastly, he recommended combining image generation with text generation. For example, he mentioned using the feature to create a personalized birthday card for a friend by generating a custom image and adding a message to it. Source : 오픈AI 제품 책임자가 밝힌 '챗GPT 잘 쓰는 법' 5가지는, AI TIMES, 21 November, 2024.
-
- 작성일 2024-12-06
- 조회수 245
-
- The Beatles' AI-Generated Song Nominated in Two Grammy Award Categories
- "The Beatles' AI-Generated Song Nominated in Two Grammy Award Categories" The Beatles are back in the Grammy race after 28 years, with their final single "Now And Then" nominated for the prestigious award. The Recording Academy, which hosts the Grammy Awards, announced that the Beatles are nominated in two categories for the 67th Grammy Awards. This nomination marks their 25th appearance in Grammy history, with the band having won seven times. Released in November of last year, "Now And Then" was completed using an unfinished tape John Lennon recorded in 1977. Paul McCartney and Ringo Starr contributed additional vocals and instruments to finalize the track. AI technology was used to enhance the audio quality, but this does not violate Grammy rules. The Grammy guidelines specify that only human creators are eligible for nominations and awards, but works that include "AI elements" may still qualify. The technology applied is similar to background noise-canceling techniques used in video chat platforms like Zoom or Google Meet. It learns specific sounds and can separate and amplify certain audio elements in a recording. Given the strong competition from artists like Beyoncé, Charli XCX, Billie Eilish, and Taylor Swift, as well as the relatively low Spotify streaming count of 78 million plays, the chances of winning are considered low, as the song’s Spotify streams are the lowest among the nominees at 78 million. Analysts believe that its chance of winning may depend on its originality. Source : 비틀즈의 AI 합성곡, 그래미상 2개 부문 후보로 노미네이트, AI TIMES, 9 November, 2024.
-
- 작성일 2024-11-14
- 조회수 256
-
- AI Emerges to Provide Real-Time Teaching Guidance to Teachers Instead of Students
- "AI Emerges to Provide Real-Time Teaching Guidance to Teachers Instead of Students" Stanford University researchers have introduced an AI-based system called 'Tutor Copilot' to improve the quality of real-time tutoring by teachers. This system does not directly interact with students. Instead, it listens to conversations between the teacher and the student and provides helpful advice to the teacher. when a teacher interacts with students, AI monitors the conversation and assists the teacher by suggesting follow-up questions or providing appropriate hints to facilitate the flow of the lesson. The main goal of Tutor Copilot is to assist less-experienced teachers. Notably, this system is based on an AI model trained on the 'think-aloud protocols' of experienced teachers, helping teachers guide students' reasoning and analyze complex concepts. This allows even novice teachers to utilize advanced teaching strategies, which is particularly noteworthy. According to a study, teachers using Tutor Copilot saw a 4% improvement in their students' understanding of math concepts. For less-experienced teachers, this improvement was even higher at 9%. This has demonstrated the effectiveness of AI in complementing the capabilities of educators. However, it has been noted that the system sometimes made suggestions that were too advanced for the student's grade level, Nevertheless, Tutor Copilot has shown that it can improve teaching performance and learning outcomes, especially in areas where educational resources are limited. This study, which demonstrated that AI-human collaboration has the potential to raise the quality of education significantly, is expected to make a significant contribution to expanding the potential use of AI in the education field in the future. Source : 학생 대신 교사에 실시간으로 교수법 조언하는 AI 등장, AI TIMES, 9 October, 2024
-
- 작성일 2024-10-10
- 조회수 290
-
- One Question to ChatGPT Costs ‘A Bottle of Water'
- "One Question to ChatGPT Costs ‘A Bottle of Water" AI services are convenient, but it has been pointed out that they require a significant amount of water. During the AI development phase, large data centers operate to train AI, with many high-performance computers running 24hours. To cool these computers, a massive amount of cooling water is needed. In fact, large data centers are said to use millions of liters of water every day. Additionally, even when using AI, such as AI chatbots, a considerable amount of water is consumed, leading to ongoing water usage. This scale of water consumption is expected to increase as AI becomes more integrated into everyday life. Recent research suggests that meeting global demand for AI services could require 4.2 to 6.6 trillion liters of water annually by 2027. Therefore, there are calls for cooperation among governments, companies, and consumers to create a sustainable future where AI development and water resource conservation are balanced. Specifically, it is suggested that governments should establish regulations and guidelines for water use in the AI industry, companies should disclose their water usage related to AI and raise awareness, and consumers should be aware of the water consumption associated with using AI services and make responsible choices. In conclusion, the perspective that everyone shares responsibility for recognizing and managing the water risks hidden behind the benefits of AI is emphasized. While AI technology will continue to advance and offer more convenience in our lives, it is also highlighted that we must recognize the environmental burdens behind this convenience and strive to reduce them. Source : 챗GPT에 질문하려면 ‘생수 한 병, mydaily, 2 September, 2024
-
- 작성일 2024-09-13
- 조회수 368