AI Agents 1(a) - What are AI Agents, and why do they matter?

AI Agents 1(a): What are AI Agents, and why do they matter? This session defines AI agents; shows where plain LLM prompting breaks; formalizes agents as LLM plus planning plus tools; and previews how we will study them in CSE 491 at Michigan State University. What you’ll learn: 1. Brief history of AI modalities: rule-based; machine learning; deep learning; generative AI; what humans vs machines provide at each step. 2. Why LLM-only systems fail at certain tasks: gaps in training data; hallucinations; model specialization; context limits near 10M tokens. 3. Agent definition: LLM with a control loop plus tools and memory; observe; plan; act; reflect; update 5. Current capability estimates: many basic technical tasks succeed at about 80 percent; some advanced tasks near 50 percent; evaluation uses task success rate, time to completion, human-intervention rate, and cost per task