Gemini Canvas
1. Gemini Canvas
2. Working with Gems
Prompting and working with AI is an iterative process. Designing, testing, and improving. Let's test this approach. Let's start with a basic prompt and use the Gemini Canvas feature to improve it. I want Gemini to generate a project plan with the milestones for planning the end-of-the-year conference for key clients. I click Canvas; it is an interactive working space in Gemini. And we get the first output. It's okay, but it's not great. Now, instead of writing a whole new prompt, I can use Canvas to interactively edit it. I can click right into the text and change whatever I think should be changed. For example, the milestones. I can select it here, and I can ask Gemini to edit it. And I can in Gemini Canvas iterate with the whole output or just with some elements until I am satisfied with the output. And this is the key workflow: prompt, review, and then refine directly on the output until you're satisfied. The Canvas feature is also great for learning. It can help you visualize complex materials. Let's test it. Here I'm uploading a dense document: Google report and prompting. Now I'll start with a simple prompt. "What are key insights?" Here, again, I select Canvas and a Thinking Model as this task is a bit more complicated. It can take a while, so we will speed it up. And now I want to create a visual representation of this document. In order to do that, I click in the bottom-right corner, Infographics. We wait, and here is the result. Just look at that! It read the document and visually structured the key concepts for me. You can actually leverage this with any document that you have, and you would like to understand it in a more visual way. What else can you do with Canvas? We can use it for small "vibe coding". We can ask Gemini to create an interactive learning tool for prompting based on this PDF. We use the same PDF that we used in the previous example. It will take a minute, so we'll speed it up. With this tool, you can dive further into prompting mechanisms and play with the results. You can actually play interactively in this learning process. It's quite cool. You can leverage this for any PDF, any dense material that you have. You can access the tool in the course description, or you can generate one on your own. So you can leverage Canvas to visually represent any dense PDF or transform any material into an interactive tool. Test it on your dense documents or maybe quarterly reports. What we did here with the report, meaning uploading the document, is a form of context engineering. To be precise, we use RAG: Retrieval Augmented Generation. You augmented the AI's general knowledge with your specific data. This is how you make AI an expert on your project, not only on the publicly available information on the Internet. And as we know, this is critical if you want to be more satisfied with the results. So, how else can you prompt and define context? I will share one last method. It's called few-shot prompting. This is actually "show, don't tell". You can give an AI two or three examples, the so-called "few shots" of exact examples of what you're looking for. For instance, you can show it two great emails you wrote and prompt the AI to generate a new email that you want that will follow the style and structure of the examples, because AI learns the pattern and style instantly. And this technique can be very effective if you have good examples of what you're looking for. And there are many other methods of prompting, like prompt chaining, tree of thoughts, etc., but they are already widely covered on the DataCamp platform. So I won't get into details here. For us in this course, the most important words for prompting are context, examples, and interaction. In our next video, we'll explore Gemini Deep Research.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.