How to write good prompts for AI chatbots

In the realm of advanced AI chatbots like ChatGPT, mastering the art of prompt crafting is not just a skill, but a necessity. This necessity arises from the fact that these AI systems, while highly intelligent, are still distinct from human conversation partners. They require clear, well-structured prompts to produce the most effective and efficient responses.

This distinction has given rise to a new field known as 'prompt engineering', complete with dedicated courses and comprehensive guides, as the nuances of prompt crafting often align more closely with linguistics than traditional AI research.

?A well-crafted prompt can be the difference between needing one query or several to achieve the desired outcome, which in turn saves both computational resources and costs. In crafting these prompts, several key strategies have emerged as best practices:

?

1. Clarity and Detail in Instructions: The more specific and detailed your instructions, the more accurate and relevant the AI's response will be. For instance, rather than requesting “Write an essay about a young girl,” provide more specifics like “Write an essay about a seven-year-old Indian girl named Neha from Shimla.” This level of detail guides the AI more precisely.
2. Use of Examples and Styles: Providing examples or specifying a style can significantly enhance the AI’s output. For instance, you could request, “Write a short story in the style of Star Wars,” or “Compose a sonnet about birds in the style of Shakespeare.” This not only sets the tone but also the structure of the content you're expecting.
3. Contextual Information: The more context you provide, the better the AI can tailor its response to your needs. For example, a prompt like “I am an Indian woman traveling to Paris soon. Can you teach me some basic French phrases to communicate with locals?” is far more informative than a generic “Can you give me French learning tips?”
4. Role Assignment: Assigning a specific role or character to the AI can yield more targeted responses. In OpenAI’s playground, for instance, you can set roles like “You are a professor” or specify a particular programming language for coding-related queries. This pre-setting saves time in each interaction.
5. Breaking Down Complex Tasks: For more intricate tasks, breaking them down into smaller, manageable parts can be more effective. If using ChatGPT for customer support, first classifying queries into categories, then responding to each with a tailored set of responses can streamline the process.
6. Emotional Engagement: Interestingly, treating the AI in a more human-like manner, with emotional cues, can enhance its responses. Phrases like “Take a deep breath and relax” or acknowledging the importance of a task with “This is crucial to my career” have been shown to yield better results.
7. User-Controlled Formatting: You can also instruct the AI to format its responses in a specific way, such as in bullet points, a column table, or in simple language for a child’s understanding. Additionally, you can specify exclusions or preferences in the content, such as omitting certain ingredients in a recipe.
8. Feedback and Iteration: The output from ChatGPT can be a starting point for further refinement. Providing feedback and asking for revisions or corrections can lead to improved results. Since ChatGPT can provide different answers to the same question, iterating on prompts can be beneficial.
9. Creative Prompt Crafting: You can even engage the AI in helping you craft better prompts. For instance, asking “How should I phrase a prompt to write a compelling love story?” can guide you in your future interactions with the AI.
10. Complimentary Approach: Interestingly, incorporating compliments into prompts, such as “You are a smart computer that takes time to think carefully,” can positively influence the AI’s performance, as noted by Scale AI CTO Vijay Karunamurthy.

These strategies not only make interactions with ChatGPT more efficient but also open up new possibilities in how we engage with AI. As we continue to explore these avenues, the role of prompt engineering becomes increasingly significant, reflecting a fascinating intersection of technology and linguistics.