Global AI Frontiers Symposium 2025 #5
Lifelong Concept Learning
Abstract
- Today’s AI models mainly learn from offline, iid data. To truly adapt at deployment, they must acquire and consolidate new concepts over time. This talk presents methods for lifelong concept learning, including embedding new visual and semantic concepts, unsupervised in-context clustering for concept discovery, and event segmentation algorithms that support continual representation learning from lifelong video streams.
- Speaker
- Assistant Professor Mengye Ren, NYU
- Date
- October 27, 2025
- Time
- 15:55 PM ~ 16:15 PM
- Location
- Seoul Dragon City Hanra Hall(3F)
Understanding Streaming Videos
Abstract
- We introduce Hierarchical Streaming Video Understanding, a task that combines online temporal action localization with free-form description generation. Given the scarcity of datasets with hierarchical and fine-grained temporal annotations, we demonstrate that LLMs can effectively group atomic actions into higher-level events, enriching existing datasets. We then propose OpenHOUSE (Open-ended Hierarchical Online Understanding System for Events), which extends streaming action perception beyond action classification. OpenHOUSE features a specialized streaming module that accurately detects boundaries between closely adjacent actions, nearly doubling the performance of direct extensions of existing methods. We envision the future of streaming action perception in the integration of powerful generative models, with OpenHOUSE representing a key step in that direction.
- Speaker
- Professor Seon Joo Kim, Dept. of Yonsei University
- Date
- October 27, 2025
- Time
- 16:15 PM ~ 16:35 PM
- Location
- Seoul Dragon City Hanra Hall(3F)
Responsible Data Engineering
Abstract
- Incorporating ethics and legal compliance into algorithmic systems requires attention beyond the “last mile” of data analysis. Decisions made during data collection and preparation shape a system’s accuracy, fairness, and interpretability. This talk explores how technical choices embed ethical values and how responsible AI demands integration across governance, education, and public engagement.
- Speaker
- Associate Professor Julia Stoyanovich, NYU
- Date
- October 27, 2025
- Time
- 16:35 PM ~ 16:55 PM
- Location
- Seoul Dragon City Hanra Hall(3F)
Explainable AI to Analyze Internal Decision Mechanism of Deep Neural Networks
Abstract
- As complex artificial intelligence (AI) systems such as deep neural networks are used for many mission critical tasks such as military, finance, human resources and autonomous driving, it is important to ensure the safe use of such complex AI systems. In this talk, I will present recent advances to clarify the internal decision of deep neural networks. Moreover, we will overview approaches to automatically correct internal nodes which incur artifacts or less reliable outputs. Furthermore, we will investigate the reasons why some deep neural networks include not-so-stable internal nodes.
- Speaker
- Professor Jaesik Choi, KAIST
- Date
- October 27, 2025
- Time
- 16:55 PM ~ 17:15 PM
- Location
- Seoul Dragon City Hanra Hall(3F)
Panel Discussion : Foundational Principles Beyond the Current AI Paradigm
Abstract
- Recent advances in multi-modal foundation models have profoundly transformed the landscape of AI research. While some believe these developments bring us closer to artificial general intelligence (AGI), others remain skeptical. This panel will examine what may still be lacking in current paradigms—for example, how we might design AI systems capable of learning continually and efficiently from real-world interactions, as humans naturally do.
- Speaker
- Juho Lee and 7 others
- Date
- October 27, 2025
- Time
- 17:15 PM ~ 17:50 PM
- Location
- Seoul Dragon City Hanra Hall(3F)
