Deconstructing the AI Black Box and Mastering the Underlying Logic of Large Language Models
Unlike general courses that heavily focus on application operations, the core framework of this workshop was based on "80% LLM basic principles and 20% privacy and ethics". During the workshop, Assistant Professor Ya-Ping Hu led the teachers in "deconstructing the AI engine," explaining the Transformer architecture and the self-attention mechanism at the foundation of AI in simple terms. The course detailed the "Tokenization mechanism," which is the basic unit for AI models to process text, as well as the fundamental nature of "Next-token Prediction" that functions like a word-association game.
From RAG to Digital Employees: Comprehensively Upgrading Teaching and Research Tools
To solve AI's "amnesia" problem and the limitations of processing ultra-long texts, the workshop further introduced the concept of the Context Window and the enterprise-level solution—RAG (Retrieval-Augmented Generation) technology. Assistant Professor Hu not only taught the teachers how to write precise Prompts, but also shared forward-looking insights on how to break AI's information silos through external Tools and the unified MCP (Model Context Protocol) standard. Ultimately, the course guided teachers on how to build "digital employees (Agents)" with autonomous planning capabilities and establish exclusive SOP skills (Agent Skills) for them.
Emphasizing Information Security and Academic Ethics, Promoting the Upgrade of System Architecture Thinking
Behind the powerful AI technology, the workshop also dedicated 20% of its proportion to focus on privacy and ethical issues. This covered AI data privacy, the security of business management and teaching data, as well as academic ethics and copyright regulations, ensuring equal emphasis on technological application and compliance.
Assistant Professor Ya-Ping Hu emphasized that the core objective of this workshop is to promote a transformation in teachers' thinking: "By bringing 'AI underlying architecture thinking' into the teaching and research of management, teachers will no longer just be consumers of AI, but efficient managers directing AI to work for you". Through the course, the goal is to help teachers shift from "passive reliance (blind users)"—treating AI as a black box and giving commands based on intuition—to "active configuration (system managers)" who can precisely write Prompts, design Contexts, and plan Agent Skills, thereby optimizing teaching and research workflows and enhancing students' critical thinking.
This workshop featured small-class sizes for in-depth exchanges. The College of Management and Social Sciences hopes that through this advanced training, it will continue to empower all teachers in the college, leading business and management education into a new era of highly efficient AI collaboration.
Photo Caption: The first session of the Advanced AI Literacy Enhancement Workshop for College of Management teachers on March 19.
Photo Caption: The second session of the Advanced AI Literacy Enhancement Workshop for College of Management teachers on March 27.
Photo Caption: Participating teachers listening attentively to Assistant Professor Ya-Ping Hu's explanation.
