From left, Hong Wan, KAIST doctoral student, and Professor Choi Moon-jung. Provided by KAIST
■ KAIST announced on the 28th that a research team led by Professor Choi Moon-jung at the Graduate School of Science and Technology Policy has quantitatively analyzed that the responses of the generative artificial intelligence model “ChatGPT-4o” embed subtle age stereotypes toward older adults. The findings were published in the February 2026 special issue of the international gerontology journal The Gerontologist. The team collected 900 texts by using neutral prompts that asked the model to describe the characteristics of age groups in 10-year intervals from 10 to 90 years old, and applied the Stereotype Content Model (SCM) to analyze the dimensions of “warmth” and “competence.” The analysis showed that for groups aged 60 and older, scores for warmth—such as kindness and consideration—were high, while competence—such as ability and expertise—tended to be expressed at lower levels than for younger age groups. In addition, as age increased, expressions related to assertiveness, which reflect confidence and leadership, decreased, indicating that AI can portray older adults as wise and benevolent while depicting them as having lower agency and proactiveness. The research team pointed out that repeated exposure to such expressions may reinforce social prejudice against older adults and lead to digital ageism.
■ Ulsan National Institute of Science and Technology (UNIST) announced on the 26th that a research team led by Professor Yoon Seonghwan at the Graduate School of Artificial Intelligence has identified, from the perspective of loss landscape flattening, why multimodal artificial intelligence (AI) learns more accurately and stably than single-modal AI. The results will be presented at the International Conference on Machine Learning (ICML 2026), a leading international conference in AI. Multimodal learning is a method in which AI uses different forms of data together—such as images, speech, and text—to better understand the same object or situation. The team explained that when data from multiple modalities are learned jointly, the loss landscape becomes flatter, thereby enhancing robustness so that the model can respond stably even to situations it has not encountered during training. They also mathematically introduced a “convolutional smoothing effect,” in which different data types mitigate rough variations in error as if averaging them out, and proposed distribution-based multimodal learning (DML), which re-pairs different modality data within the same target class. Researcher Lee Jaejun of the UNIST Graduate School of Artificial Intelligence participated as first author, and the team reported that DML outperformed conventional fixed-pair learning in image–text retrieval and classification experiments.
■ Gwangju Institute of Science and Technology (GIST) announced on the 28th that it held the “GIST Integration and Innovation Forum” at the GIST Oryong Hall on June 26 and presented a vision for a science and technology-based innovation ecosystem in preparation for the era of the integrated Jeonnam–Gwangju Special City. The forum brought together about 100 participants from GIST and industry, academia, research institutes, and government to discuss regional growth strategies for the age of AI transformation (AX) and wide-area integration. On the same day, GIST also held the opening ceremony for the Startup Innovation Promotion Center and signed a memorandum of understanding with the Chonnam National University Industry-Academic Cooperation Foundation and the KENTECH Value Creation Center of the Korea Institute of Energy Technology (KENTECH) to promote regional innovation, technology startups, and commercialization. The Startup Innovation Promotion Center will operate as a dedicated organization that provides integrated support for commercialization of research成果, technology startups, investment linkage, and industry–academia cooperation. At the forum, Professor Kim Jae-kwan of the Department of Biomedical Science and Engineering, Dean Kim Kanguk of the School of Information and Communications, Professor Park Kihong of the Department of Environmental and Energy Engineering, and Special Professor Kwak Jaewon of the AI Graduate School of Policy and Strategy gave presentations on topics such as AI semiconductors, energy, deep-tech startups, and visions for the coexistence of humans, AI, and the Earth. GIST President Lim Ki-cheol stated, “By using the Startup Innovation Promotion Center as a hub to connect research, startups, and industry, we will take the lead in enabling Gwangju and Jeonnam to emerge as a core driver of Korea’s future growth.”
– doi.org/10.1093/geront/gnaf291
Copyright ⓒ DongA Science. All rights reserved.