A guide to writing effective AI prompts (current through 2024): Leverage roles, actions, context, and formats

In April 2024, Google released a document outlining how to effectively utilize its AI, Gemini, as shown in the image below. Highlights from the document include 4 guides to writing effective promptsin the previous post. In this post, we're going to talk about that.

* This article is based on Google's publicly available documentation, but there may be some errors, so please refer to the original documentation to understand the exact meaning.

The post One-line summaries

First, to summarize Effective prompting requires four elements: Persona, Task, Context, and Structure/Format.in the file.

While this handbook is not specific to Gemini, it can be applied to any AI service, so let's dive into the details of each element.

4 AI prompting essentials

1. Persona (Assign roles)

Gaslighting the AI about what roles and perspectives it should have

For example, "You're the Chief Marketing Officer (CMO) of a large pharmaceutical company, and you're about to launch a new diabetes drug, and you need to strategize a marketing campaign for this new product."

By specifying these personas, the AI will respond with a CMO perspective and tone of voice. It uses knowledge specific to that role, such as product and customer characteristics and marketing theory, to create more actionable outputs.

2. select Task

In this step, you tell the AI a specific task to perform

For example, "Summarize," "Write," "Rephrase," etc. More specifically, "Based on previous product launch performance data and customer research, please provide me with three key directions for the marketing strategy for the new diabetes drug."

For complex tasks, break the task down into steps and give each step a specific task to perform. This way, the AI understands the overall task and is better able to respond to your requirements. In other words, be specific about the purpose and deliverables of the task.

3. click Context

The context phase is about giving the AI the relevant background and details it needs to fulfill your requirements.

For example, "Use the latest research reports and customer database data on diabetes treatment to inform your strategy. You can find relevant information here. [URL 1], [URL 2], [URL 3]." At this stage, it's also a good idea to tell us about your users so we can understand who this product or service is targeted at.

By providing additional context, such as relevant documents, data, and situations, AI can produce more relevant results.

4. click Format

In this step, specify the desired output format

For example, "2 paragraphs of an executive summary", "a list of 10 ideas", etc. More specifically, you might say, "This should be written in the form of a marketing strategy document, with a professional but layman's tone. The document should include an executive summary, an analysis of the current state of the market, and a description of the three key strategies, and the entire document should be around 800 words."

The more specific formatting instructions you give, such as character limits, language style, and post structure, the more likely you are to get the results you want.

Iterate and improve prompts

It would be nice to end up with 1 through 4, but as we all know, one or two prompts won't get you the results you want. It's important to keep the conversation going with the AI by repeating prompts and making small tweaks.in the file.

You'll get better results if you iterate through feedback as you gradually refine your prompts and review the AI's responses. For example, you might add instructions to the initial prompt, such as "Add more details" or "Change the tone to formal" or "I'm happy with the draft so far, but I think the introduction is a bit bland. Can you flesh out your opening sentence to make it more engaging for the viewer?"

This iterative process of refining our prompts and feeding them back and forth to the AI gets us closer to the results we want. To summarize, prompting is never perfect the first time, so iterative refinement is essential, and this is how we can improve the quality of the AI's output.

Organizing prompt writing

In fact, if you look into the content, if you are using AI services frequently, you are already doing everything, but I didn't know that it was named like that. I'm already doing it, but I wonder if I can make more effective results if I keep the above content in my head once more.

To get the most out of your AI assistant, it's essential to craft effective prompts that take into account role assignment, task, context, and structure/format. By assigning personas to prompts, providing specific work instructions and relevant context, and talking about the desired output format, AI is more likely to produce appropriate results.

Also, because prompting is a dialog with the AI, one prompt may not produce the desired result, but rather a process of iteration, refinement, and interaction with the AI. This incremental refinement of prompts and feedback will result in more refined and polished results.

For reference, see the previous post GPT-4 is free? (Feat. AI Service Alan)You can try the above using Alan, which we introduced in the article.

We're cheering you on and applauding you for leveling up your IT knowledge with this post.

All content on 'secondlife.lol' is protected by copyright law. Unauthorized reprinting, copying, distribution, etc. is prohibited.