All Sessions：

In this session, we will explore the world of large language models (LLMs) from a hacker's perspective. By deploying Ollama locally as an attacker, we will simulate automated attacks targeting various systems using LLMs. The live demonstration aims to highlight the critical importance of LLM security. Participants will experience firsthand how LLMs can be weaponized and will engage in discussions on strategies to defend against such threats. This will emphasize the need for robust security measures in AI-driven environments. Using HackingBuddyGPT as a prototype, combined with Ollama deployed locally, we will demonstrate automated attacks on target systems. Regardless of whether the attacks succeed, this research will reveal how LLMs, while offering convenience, can also be exploited as tools for malicious hacking. This serves as a warning to developers, researchers, and others, raising awareness of the importance of LLM security.

Latest Applications of Large Language Models: AI Fortune Interpretation, AI Content Creation, AI Oracle Explanation, AI Scripture Interpretation, AI Customer Service Latest Applications of Image Analysis Algorithms: AI People Counting System, AI Guardian System (Fire Detection, Smoke Detection, Anti-theft) 最新大型語言模型應用：AI 解籤、AI 小編、AI 解鸞文、AI 解經文、AI 客服 最新影像分析演算法應用：AI 人流系統、AI 門神系統 (防火、防煙、防盜)

Non-programmers are now using AI to write code and automate tasks that make their daily work easier. I've been running study groups ([blog.gaiconf.com 連結]) to help them learn how to do this. In my talk, I want to share some of the incredible things these non-engineers have been able to build with the help of AI.

Multimodal AI is revolutionizing traditional manufacturing by equipping manufacturing equipment with stronger perception and decision-making abilities. This enables automation and intelligence in production processes, allowing enterprises to create more resilient and efficient smart factories. 隨著多模態 AI 技術的成熟，傳統製造業正在經歷一場深刻的轉型。多模態 AI 能夠賦予製造設備更強的感知能力和決策能力，實現生產過程的自動化和智能化。透過多模態 AI，企業能夠打造更具彈性、更具效率的智慧工廠，從而提升市場競爭力。

Exploring how AI-driven visual inspection platforms and generative image technology enable fast and high-precision defect detection, while boosting production efficiency. This session will focus on the role of these cutting-edge technologies in accelerating digital transformation within the manufacturing industry. 探討如何透過 AI 視覺檢測平台與生成圖像實現快速與高精度瑕疵檢測和生產效率提升，助力製造業數位轉型

Eval-Driven-Development (EDD) is a method for developing LLM-based apps guided by task-specific evaluations, including dataset synthesis, evaluation metric design, and automatic prompt optimization. In this short talk, I will quickly demonstrate this process.

The presentation will focus on the principles of multimodal language model edge Computing technology and its applications in healthcare, demonstrating its extensive use in six different medical scenarios. 1. Chronic disease management Q&A system for dementia knowledge : web-based platform with real-time, accurate medical information 2. Line Chatbot for remote diabetes care : facilitates intelligent Q&A between patients and healthcare providers. 3. Emotional companion QA robot : TVBS interview video demonstrating psychological support via language models. 4. Digital interaction display for standardized dental-patient interactions : AI automates assessments for effective dental education. 5. X-ray dental image report generation system : automatically generates dental imaging reports to improve medical efficiency. 6. Text-to-dental 3D image generation technology : converts text into 3D images for VR imaging and surgical planning in dental surgery.

The intersection of LLM and privacy-preserving techniques is an emerging research field, particularly relevant for domain-specific applications. There is increasing interest in methodologies that enable collaborative or multi-tenant training and inference while addressing both trust and efficiency concerns. In this talk, I will give a high-level tutorial on several cryptographic techniques that may help the audience better understand the state-of-the-art research in the area of secure AI inference.

Dr. Lee-Feng Chien has long been focusing on the development of software technology and innovation. From 2006 to 2020, he served as the first Managing Director of Google Taiwan and successfully transformed it into the largest R&D center in Asia. Post-retirement in 2020, Dr. Chien has been actively engaged in Taiwan's entrepreneurial ecosystem, serving on the boards of AI startups such as Appier and iKala. Notably, Appier was successfully listed in Japan and became Taiwan’s first AI unicorn. He also co-founded an angel fund, and mentors numerous startups, as well as serving as a member of the Executive Yuan's Economic Development Committee. Dr. Chien is also deeply involved in academia. As a pioneer in Chinese natural language processing and AI, he obtained his Ph.D. from NTU CSIE in 1991 for his research on Chinese language models. Between 1993 and 2006, he was a research fellow at the Institute of Information Science, Academia Sinica, while concurrently teaching at the Department of Information Management at National Taiwan University from 2001 to 2006.