Prompting Gemini effectively
1. Prompting Gemini effectively
2. Prompting Gemini effectively
Welcome back. In a previous exercise, you identified your repetitive and high-value tasks. Now, we are going to learn how to use our AI assistant to tackle them. It starts with one of the oldest rules in computing, which you are probably aware of. It's called GIGO, which means that the quality of the output is very much dependent on the quality of what you put in. If you give Gemini a lazy, vague command, you will likely get a lazy, vague result. Now, this rule has exceptions. If you're doing a simple mechanical task like "Translate this sentence into Spanish" or "Fix the grammar in this paragraph," a short, direct prompt is perfectly fine and efficient. But for any of your high-value creation tasks, good input with clear context is critical. So what makes a quality prompt? Context. Let's see how it works. Let's ask Gemini to generate an image of a watch. No need to prompt more. Just generate an image of a watch. You can now pause the video and do it in another tab. Ready? Pause. Okay, if you paused and now resumed. Welcome back. Now, my guess is that your image of a watch is showing the same time as my image of a watch. Meaning between 10:08 and 10:11, right? How do I know it? Let's look at the data in Google search. I type "watch" and this is what I see. Most watches show the same time. Why? Because someone, at some point, decided that watches look the best when they show 10:10. What is the conclusion here? If you use general prompts, you'll get the most probable generic results. And we all want to do it better, right? So let's test it on a work example, on writing an email. We'll go from the level "basic" in prompting to level "great" in prompting. Let's start, of course, with the level basic. The prompt will be as following: "Write me an email to my boss about the need for the next year planning." Let's see the result. It's okay. You can work with that. But without edits, I wouldn't probably send it because it's kind of generic. Why? Because the AI doesn't know the project, the boss's name, or the expected communication style. For example, the boss might prefer very short emails. So let's move to level "better". We can get much better results by adding simple details. So in this example, I asked Gemini to write a professional email to my boss Jane about starting next year's planning. I want to mention our team's success with project Phoenix as a reason to plan ahead, and I want to keep it under 100 words. Let's see the results. It looks slightly better and more precise because the prompt now is better. Why? One, it uses contextual prompting. We provided specific details like project Phoenix or the boss's name. Two, the prompt is specific about the output. We added a clear constraint: write under 100 words, because, for example, our boss may prefer this type of communication. So now let's go to the level great. Here, we give the AI a complete picture to draft a strategic email. I'm pasting the prompt that I prepared before, where I defined role, context, and task. Importantly, I specified key learnings for the next year to be implemented. Let's see the result. There are minor differences, but thanks to those minor differences, we would spend less time on editing the output. And this is what we expect from AI. We don't want to spend hours to just refine the output. We want to get it right as quickly as possible. So, what were the prompting techniques that we used in this example? First role prompting: act as a senior project manager. This sets the tone and expertise level. Second rich contextual prompting. We provided the boss's name, title, and personality: data-driven, and we specified project details. Third, it provides clear instructions: The four-point list clearly defines what the output must contain, which is a powerful way to guide the AI in any of your examples. So, the takeaway from this video is the following. Prompts should be concise, clear, and easy to understand for both the user and AI. Second, you should be specific about the output. Providing specific details in the prompt can help the model focus on what is relevant, and thanks to that, improve overall accuracy. Final thought, think of prompts as a brief for an AI. A great brief is just the start of the conversation. And of course, the better the brief, the better the output. But you don't have to accept the first answer. You should iterate over the result. And we'll cover how to do it in the next video. See you there!3. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.