生成式AI技術與應用

臺北科技大學 114-2 學期課程，課號 354774。

課程大綱

The rapid development of Generative Artificial Intelligence (GAI) has profoundly transformed content creation, interaction design, and information system development. This course adopts a practice-oriented approach, focusing on the application and system integration of generative AI in real-world projects. Emphasis is placed on programmatic interaction with generative AI through APIs, enabling students to transform generative capabilities into controllable, reproducible, and integrable system functions. Following a realistic project development workflow, the course begins with Google AI Studio to support rapid prototyping and prompt design, allowing students to understand the characteristics of text, image, and multimodal content generation. The course then moves to Python-based implementation using the Gemini API, where students develop modular and reusable generative AI services suitable for system integration and deployment. In the later stage, Ollama and LangChain are introduced to explore hybrid architectures that combine local and cloud-based language models, highlighting practical considerations such as system design, performance, cost, and data privacy. This course adopts a project-based learning approach in which students continuously develop and refine a single project throughout the semester. Weekly deliverables are incrementally integrated into the final outcome, which is expected to demonstrate real-world applicability, technical completeness, and presentation readiness. Through this process, students will also address key issues including generation quality, cost control, ethical considerations, and commercial feasibility, thereby developing the ability to effectively deploy generative AI technologies within information systems and digital services.

課程進度

Week 1 (Feb 23): Course introduction and project topic definition (overview of generative AI projects, team formation, and application scenarios) Week 2 (Mar 2): Google AI Studio and prompt design fundamentals (text generation) Week 3 (Mar 9): Google AI Studio for multimodal applications (text and image generation with Imagen) Week 4 (Mar 16): Application scenarios and generative AI service prototyping using Google AI Studio Week 5 (Mar 23): Python integration with the Gemini API (text generation module) Week 6 (Mar 30): Python integration with the Imagen API (image generation module) Week 7 (Apr 6): Tomb-Sweeping Day make-up holiday (no class) Week 8 (Apr 13): Python-based multimodal integration (text-driven image generation pipeline) Week 9 (Apr 20): Midterm project integration and presentation Week 10 (Apr 27): Ollama and local large language model implementation Week 11 (May 4): LangChain fundamentals and generative AI workflow design Week 12 (May 11): LangChain with RAG (project data integration and retrieval-augmented generation) Week 13 (May 18): Hybrid architecture design with cloud and local models (Gemini and Ollama) Week 14 (May 25): Generation quality, ethics, commercial feasibility, testing, cost control, and system stability Week 15 (Jun 1): Project integration and presentation interface design Week 16 (Jun 8): Project optimization, testing, and stability improvements Week 17 (Jun 15): Final project preparation and integration Week 18 (Jun 22): Final project presentation

評量方式

1. Weekly Project Deliverables (40%) Students are required to complete weekly project outputs aligned with the course schedule, including prompt design, generative AI API integration, implementation of generation modules, multimodal integration, and system workflow construction. These deliverables are accumulated as the foundation of the final project. 2. Midterm Project Integration (20%) The midterm assessment focuses on project integration. Students present the current status of their project, demonstrating implemented system functions, generative workflows, and technical approaches through presentations or demonstrations. 3. Final Project Outcome (30%) The final assessment evaluates the completed project in terms of functional completeness, quality of generated content, level of system integration, and real-world applicability. A final presentation and demonstration are required. 4. Class Participation (10%) Students are assessed based on attendance, participation in hands-on activities, engagement in class discussions, and constructive feedback provided to peers.

教材

【遵守智慧財產權觀念，請使用正版教科書，不得使用非法影印教科書】 使用外文原文書：否 Use of self-made material Course videos link url: https://drive.google.com/drive/folders/1gX3poY1stLeTbbGCmIfLqB4t5UxTQVXt?usp=drive_link (Requires an NTUT Gmail account to log in.)