In-context learning for ChatGPT: Enhancing autonomous AI - ChatGPT Tutorial

- I want to take you back to that intern analogy again. And I want to think about the different ways that we teach someone to perform a task or we instruct them to perform a task. And one of the ways we do it is with explicit instructions. And we've been doing that over and over. We're telling the agentic AI, "Here's what you're going to go and do. You're going to do it step by step." But one of the ways that we teach, for example, interns in the real world is we go and show them examples of the problem being solved correctly. So if we want them to write a report, we say, "Here's the report, here's my thought process on how I put together this report, here's the final result." Or if we want them to go and diagnose an issue in software, if they're a developer, we say, "Here was the feature I was trying to implement. Here's how I walked through debugging it or building it or designing it," whatever that thought process was, here's the result. And it turns out we can do the same thing…