As a reader and a practitioner of data science and/or full-stack web development, you know your craft. Knowing your craft means you know the tools and techniques to solve problems. At this point, you’re looking at an AI assistant and realize it’s controlled by natural language, so-called prompts. What you may not realize is that there’s more to it than just writing a prompt and getting an answer. An AI assistant is trained on a large corpus of text, so it’s quite flexible on what it can generate text on and how to respond to prompts. Because of this flexibility, it’s important to understand how to write prompts that are effective and efficient.

Prompts are input to AI tools. Depending on what you’re trying to achieve, you need to adjust your prompts for the scenario you’re solving for. Therefore, how you “prompt” matters. For example, if your prompt is too vague, you won’t get what you need. Or, let’s say you’re trying to use a prompt to generate company slogans; you don’t want to use the same prompt for generating code for an app. Conversely, in a discipline like data science, it’s important you perform tasks in a certain order and your prompt should reflect what you want done and, if needed, the steps to do so.

What you need to succeed is an approach, a strategy, that you can use in general to be efficient with AI assistants. Additionally, such a strategy should be specific enough to present “best practices” for chosen problem domains. As mentioned earlier in this chapter, we’ve developed a prompt...

Let’s apply the different guidelines presented in the former section, and through using these guidelines, we will step by step work ourselves toward a prompt strategy.

Break down the problem: “web system for inventory management”

Let’s use a real example, “inventory management” within web development, to see if we can understand the general problem. To “manage” an inventory, you need to read and write data to it. Most likely, you will have different roles in this system/app, everything from administrators to normal users. You might also need to consider how this system fits in with other systems, should you, for example, integrate it with other systems, what parts it consists of in that case and how.

The domain seems pretty straightforward so let’s move on to understand what parts it consists of.

At a high level, we understand what the system should do. But to solve...

Let’s do a similar thought experiment for data science as we did for web development. We’ll use the presented guidelines “problem breakdown” and “generate prompts,” and just like in the web development section, we’ll draw some general conclusions on the domain and present those as a prompt strategy for data science.

Problem breakdown: predict sales

Let’s say we’re building a machine-learning model to predict sales. At a high level, we understand what the system should do. To solve the problem though, we need to divide it into smaller parts, which in data science usually entails the following components:

• Data: The data is the part of the system that stores information. The data can come from many places like databases, web endpoints, static files, and more.
• Model: The model is responsible for learning from the data and producing a prediction that’s as accurate...

The most important part of this strategy is verifying correctness and that the text and code created by the AI assistant are correct. There are two general approaches we can take to verify our outcome:

• Verification via prompts: The first approach is to use prompts to verify the outcome. This means writing prompts that question the outcome of specific results. This can be a good strategy to employ at the beginning of your verification process. What you’re looking for are situations where the AI assistant isn’t consistent in its responses.
• Classical verifications: The second approach is to use classical verification techniques. What those techniques are varies depending on the problem domain. At a high level, though, it boils down to testing code, comparing output, and relying on your own knowledge, and the knowledge of your peers, to verify the outcome.

The AI tool doesn’t really know what it’s doing. The...

Throughout this chapter, we’ve provided a strategy for solving problems with prompts and validating the solution.

You’ve seen how both web development and data science can be broken down into smaller parts that can be solved with prompts. We also identified some basic principles for writing prompts.

Finally, we looked at how to validate the solution using prompts and classical verification techniques.

It’s our hope that you will revisit this chapter when you’re looking at solving a problem within web development or data science and you’re looking for an approach.

There’s more to prompting than writing a prompt and getting a response. You will see throughout this book how we use these principles in various domains to solve problems. Try typing these prompts as you read, adapt to your own needs, and see what happens.

In the next chapter, we are going to learn more about the two AI assistants of our choice, GitHub...