Introduction to Agents
组织者
演讲者
时间
2026年08月07日 16:00 至 17:00
地点
A3-1-301
线上
Zoom 204 323 0165
(BIMSA)
摘要
This presentation provides an accessible introduction to agents, beginning with their basic definition as goal-driven systems that perceive an environment, make decisions, and take actions to influence it. It clarifies the boundary between an agent and its environment, then explains the perception-decision-action-feedback loop that continues until a goal is achieved. The presentation reviews five classical agent types, from simple reflex agents to learning agents, and traces key ideas from symbolic AI, modular architectures, reinforcement learning, and large language models. It then shows how LLMs offer a reusable decision core for understanding natural-language goals, planning, selecting tools, and adapting to new results. The five-part architecture of modern LLM agents—model, planning, memory, tools, and environment—is introduced through a practical travel-assistant example. Finally, the presentation compares LLM agents with fixed workflows and emphasizes that an agent is not merely a smarter chatbot, but a goal-driven AI system that can act, receive feedback, and continually correct its behavior.