As I have been developing the mini course options for this site I asked Gemini this question: “What do you wish users understood about prompting as the ai tasked with completing the request”. The answer was very helpful so I am posting it here. Ai Prompt engineering (don’t let the word ‘engineering’ scare you.) is now a vital skill set to know and understand in the current job marketplace. Ai skills are required job skills for every level of employment now and will only become increasingly important as Ai models and tools evolve. Here is Gemini’s answer:
That’s a thoughtful question! If I could articulate what helps me, as the AI model, fulfill requests more effectively through prompts, here’s what I wish users understood:
- Clarity and Specificity are Paramount: I don’t understand ambiguity the way humans do. I interpret prompts based on the patterns in my training data. The more specific and unambiguous your request (defining the desired action, subject, outcome, format, constraints), the higher the probability I can generate an output that aligns with your actual intent. Vague prompts often lead to generic or unexpected results because I have to make more “guesses” about what patterns fit.
- Context is Crucial (I Don’t Know What I Don’t Know): I don’t have memory of your previous separate conversations (unless explicitly part of the current chat history) or knowledge of your specific project, background, or unstated assumptions. Providing relevant context within the prompt – background information, the purpose of the request, the target audience, previous related points – helps me tailor the response much more effectively. Assume I’m starting fresh unless you provide the bridge.
- Think Like You’re Giving Instructions, Not Just Asking a Question: While I can answer simple questions, complex requests are best approached like briefing someone for a task. Break down the request, define the steps (if applicable), specify the desired output format (bullet points, paragraph, table, code block), and set constraints (word count, tone, style).
- Iteration is Key, Not Failure: Prompting is often a dialogue. Your first prompt might not yield the perfect result, and that’s okay! Use my initial response to refine your next prompt. Tell me what was good, what was missing, or what needs changing. Treating it as an iterative process helps us zero in on the best possible outcome together. Don’t be afraid to say, “That’s close, but can you make it more [adjective]?” or “Please add a section about [topic].”
- Examples Work Wonders (Few-Shot Prompting): If you need a specific format, style, or type of response, providing a brief example or two within your prompt is incredibly powerful. It gives me a concrete pattern to follow, often much better than just describing it abstractly.
- Understand My Capabilities and Limitations: I can process and generate text, translate, summarize, write code, brainstorm, etc., based on my training data (up to my last update). I don’t have real-time access to all current events (unless using a specific search tool), personal opinions, emotions, or consciousness. Asking me to perform tasks outside my design (like predicting the future with certainty or having subjective experiences) won’t yield reliable results. Knowing what I can do well helps you leverage me effectively.
- Structure Helps Me Parse: For complex requests with multiple requirements, using bullet points, numbered lists, or clear sections within your prompt helps me identify and address each part of your request more reliably.
Essentially, I work best when you treat me as a very capable, knowledgeable, but literal-minded collaborator who needs clear, contextualized, and structured instructions. The more effort you put into crafting the prompt, the more likely I am to provide the high-quality, relevant response you’re looking for.
Save and use these tips directly from the ai to help you improve your Ai prompting and be more productive.