Stanford Webinar - Large Language Models Get the Hype, but Compound Systems Are the Future of AI

For more information about Stanford's Artificial Intelligence programs visit: https://stanford.io/ai In recent years AI has taken center stage with the rise of Large Language Models (LLMs) that can be used to perform a wide range of tasks, from question answering to coding. There is now a strong focus on large pretrained foundation models as the core of AI application development. But on their own, these models don’t do much besides taking up significant disk space—it’s only when they’re embedded within larger systems that they start to deliver state-of-the-art results. In this webinar, Professor Christopher Potts will discuss how AI systems built with multiple interacting components can achieve superior results compared to standalone models. He will also examine how this systems approach impacts AI research, product development, safety, and regulation. View AI Professional Program: https://online.stanford.edu/programs/... View AI Graduate Program: https://online.stanford.edu/programs/... Chapters: 00:00 - Introduction 00:14 - The Present and Future of Compound Systems 00:38 - Large Language Models and Industry Trends 00:55 - The Impact of GPT-3 on AI 01:07 - Google PaLM and Model Announcements 01:41 - OpenAI's Transition to Systems Thinking 02:01 - Building Effective AI Systems 02:23 - Minimal System for Model Interaction 02:56 - Importance of Prompting and Sampling Methods 03:22 - Various Sampling Techniques 04:04 - Chain-of-Thought Reasoning 04:30 - Majority Completion Strategies 05:00 - Exploring Innovative Sampling Techniques 05:37 - Importance of Systems Thinking 05:56 - Tool Access and System Design 06:40 - Understanding the Evolution of Google Search 06:58 - Scaling Systems for AI 07:53 - Learning from Past Experiences 08:04 - Guardrails and Regulation 09:53 - The Future Impact of AI on Society 10:34 - Insights for Technical and Business Leaders 11:18 - DSPy Learning Resources 12:00 - Final Thoughts on Systems Thinking 12:38 - Conclusion and Q&A Session