This page covers advanced prompting techniques specifically for ChatGPT.
11. Continuing Conversations
ChatGPT remembers the context of previous conversations. Leverage this to perform complex tasks over multiple turns or explore a topic in depth. Clearly formulate follow-up questions or additional instructions based on previous responses.
Examples:
User: "Explain the basic concepts of quantum computing." ChatGPT: (Explains quantum computing) User: "What is quantum entanglement, and how can it be utilized?"
12. Handling Long Prompts
ChatGPT can process long prompts. It's often beneficial to provide all necessary information at once to help the AI grasp the full context. Structuring the information can improve readability in such cases.
Examples:
"The following is a draft report on a recent cybersecurity breach incident. [Insert report text - thousands of words] Read this report and propose three most urgent issues to address, along with concrete solutions for each. The answer should be summarized within 500 words and written in language easily understandable by executives, rather than technical details."
13. Multi-turn Prompting
If a complex result is difficult to achieve with a single prompt, you can incrementally build the output through multiple interactions. Review the AI's response at each turn and provide instructions for the next step.
Examples:
User: "Suggest 5 names for a new coffee shop." ChatGPT: (Suggests 5 names) User: "I like 'Morning Brew' from your suggestions. Now, create three taglines for a coffee shop named 'Morning Brew'."
14. Utilizing Specific Features
If ChatGPT has access to specific features or plugins, you can explicitly mention them in your prompt to encourage the AI to use them (e.g., web browsing, code interpreter).
Examples:
"Using the web browsing feature, find the global smartphone sales statistics for the first half of 2024 and summarize the market share by major manufacturers in a table." "Execute the provided Python code and explain any potential runtime errors that might occur, along with their solutions: [Insert Python code]"
15. Error Handling & Correction Requests
When there are errors in the AI's response or areas that need improvement, you can specifically point out what is wrong and request corrections. Clear feedback helps the AI learn and improve its subsequent responses.
Examples:
User: "In your previous response, the explanation of 'nuclear fusion' seemed somewhat inaccurate. Could you explain the energy generation mechanism more precisely?" User: "The code you provided assumes a specific library version. Could you modify it to work with the latest version?"