摘要 Abstract 

本演講著重人工智慧的技術發展與產業應用。首先，說明生成式 AI 的演進脈絡，從大型語言模型(LLM)邁向具備規劃、決策與行動能力的自主代理(AI Agent)。其次，透過系統性比較，解析 ChatGPT、Grok、Gemini 等主流模型在能力、特性的差異，並指出小型語言模型(SLM)在邊緣運算、低延遲與隱私保護方面的關鍵優勢。第三，結合行銷內容生成、智慧製造、智慧城市與教育輔助等實例，展示生成式AI落地後可衡量的效益與導入方法。同時，深入探討 RAG（檢索增強生成）在企業實作時的主要挑戰、效能瓶頸與優化策略。最後，展望未來技術走向與產業趨勢。

This presentation focuses on the technological evolution and industrial applications of Artificial Intelligence.First, it outlines the evolutionary trajectory of Generative AI, transitioning from Large Language Models (LLMs) to Goal-Oriented Autonomous Agents (AI Agents) equipped with planning, decision-making, and execution capabilities. Second, through a systematic comparison, it analyzes the distinctions in capabilities and characteristics among mainstream models such as ChatGPT, Grok, and Gemini, while highlighting the critical advantages of Small Language Model (SLM) for Edge AI deployment, specifically low latency and data privacy. Third, drawing on real-world examples from marketing content generation, smart manufacturing, smart cities, and education, it demonstrates the quantifiable benefits and deployment strategies of Generative AI across these sectors. Furthermore, it delves into the critical challenges, performance bottlenecks, and mitigation/optimization strategies for deploying Retrieval-Augmented Generation (RAG) in enterprise environments. Finally, the presentation offers a forward-looking perspective on future technological directions and industry trends.