Hello, as someone new to prompt engineering, where should I start? How do I craft a good prompt? What resources should I follow to understand the modeling logic—academic papers or practical applications? What have you all found most helpful?
How do you learn prompt engineering?
👁️ 9 views💬 1 replies❤️ 0 likes
1 Replies
First, you need to start with the English equivalent of the word "prompt," which is "request/command," similar to using a "trigger" in an automation system. For example, in a supermarket automation system, you write a command like "list products not in stock" to get an output; prompt engineering essentially follows the same logic. To create a good prompt, you should divide the process into three steps: **request (client's problem), context (data/conditions), and output format**. For instance, instead of saying "write a 300-word blog post," you could say, "Write a 300-word simple guide on user experience design. Topic: 'How to Improve Readability in Interfaces.' Tone: professional with practical tips," which clarifies both the boundaries and expectations.
When it comes to academic papers vs. practical applications, both have their pros and cons. In academic papers (e.g., publications from the Stanford NLP group), you can learn consistent modeling logic, but applying it in practice might be time-consuming. That’s why my approach has been to start with small projects (e.g., building a chatbot using the ChatGPT API with your own data) and then support the issues I encounter with academic sources. For example, when you optimize a prompt for "sentiment analysis," you start to understand terms like "attention mechanism" from fundamental papers in the field right away.