Loading video player...
This video, generated by NotebookLM, serves as supplementary material for the "LLMs & AI agentic Systems" workshop at Taiwan Soochow University. Are we moving past the era of simple AI search tools? This video explores the rapid transition from standard, one-shot Retrieval-Augmented Generation (RAG) systems to multi-step autonomous deep research agents. These advanced systems function like human analysts, utilizing a continuous "plan, search, read, synthesize" loop to manage complex tasks over extended compute horizons. Watch a breakdown of how a multi-agent workflow can autonomously research the 2030 solid-state EV battery market, pivot past dead ends, use vision models to interpret charts, and run Python code to compress days of human desk research into a fully cited report in just 15 minutes. However, with great power comes significant risk. We also dive into the "RAG paradox" and the dangerous clinical vulnerabilities that emerge when agents face hallucination cascades, potentially leading to unchecked automation bias from users. Finally, discover how enterprise frameworks like Nvidia's AIQ manage these hallucination risks and high compute costs through rigid architectural routing and hybrid models. 📍 About KYC AI Labs: An innovation hub by KYC Global Pte. Ltd. (Singapore), dedicated to bridging the gap between advanced AI research, governance, and practical business application. 【Keywords】 #TaiwanSoochowUniversity #DrKuangYuChow #KYCAILabs #DeepResearchAgents #AIAgents #LLMs #AgenticSystems #NotebookLM #AutonomousAI #AIHallucinations #AutomationBias #NvidiaAIQ