How can I prompt?

In the field of AI, prompting is used to make it easier to interact with AI models. It allows us to train and control AI models by giving them specific instructions or examples. Prompting also helps improve user experience and prevent misuse of AI systems.

The main goal of prompting is to facilitate learning and performing new skills, especially for people with learning disabilities or behavioral problems.

And this is how we talk to machines. Because in the context of computers, the term “prompting” refers to the prompt that a user receives to perform a specific action or to provide an input to perform an action, whether in a user interface or in a program.

Why do we prompt with AI?

By giving specific instructions or examples, AI models can be trained or controlled to solve specific tasks or problems.

There are various reasons why we prompt with AI:

1. Training AI models: AI models must be trained with large amounts of data in order to recognize patterns and connections. Prompting makes it possible to direct the desired patterns or behaviors by giving the models specific examples. By clearly defining the desired behavior in the instructions or examples, one can train the AI models more effectively.
2. Control and adjustment: Prompting allows you to control and adjust the results and responses of the AI models. By providing specific instructions or criteria, one can ensure that the AI models produce the desired results or focus on specific requirements.
3. Improving user experience: Prompting also makes it possible to simplify user interaction with AI systems. Instead of having to formulate complex commands or requests, users can give simpler and more natural instructions. The AI models then understand and interpret these instructions to produce the desired results.
4. Avoiding misuse: Prompting can also ensure that AI systems are not misused or used for malicious purposes. By setting certain restrictions or guidelines in the instructions, one can avoid undesirable behaviors or outcomes.

It is important to note that prompting is only one aspect of interacting with AI systems and that its application depends heavily on the specific application, context and goals. However, it enables effective communication and collaboration between humans and AI models.

How do I learn to prompt?

First of all, it has to be said that prompting is not a science and many tools allow you to skip prompting because they already have templates. Still, it’s a great way to better understand how we can deal with generative AI.

1. To do this, understand the concept of prompting: Learn how prompting works and how to formulate specific instructions or examples to achieve the desired results. Read guides, tutorials, and case studies to get a better understanding.
2. Experiment with examples: Start with simple examples and try to guide the AI model with clear and concise instructions. Test different approaches and observe the output results.
3. Reflect and improve: Analyze the results of your prompting requests and try to understand why certain phrases or instructions produce better results than others. Use this knowledge to further improve your prompting skills.

And then there are free online sources such as: LearnPrompting.org)

What is a good prompt?

Very important: Prompting is a dialogue in which your counterpart, the AI, should not guess, but you are the lead and you specify the context, intention and content. The more precise and goal-oriented you are and the more content you provide, the better. A few rules include, among others:

1. Be friendly: Communication starts with politeness. Remember to use “please” and “thank you.”
2. Give context: The more information you provide, the better the AI can respond to your request. Provide background information so that the AI understands the overall context.
3. Be precise: Ask a clear and direct question or task. Use appropriate keywords, they will help the AI understand your requirement better.
4. Formulate goal-oriented instructions: Your prompt should contain an instruction that works towards a specific goal. This means the AI knows exactly what you expect from it.
5. Have a conversation: Imagine you are having a conversation. With prompts, you are in charge.
6. Use simple language: Make your sentences short and understandable. Avoid complex sentence structures.
7. Ask questions: If you receive a response from ChatGPT, ask additional questions to get more details. Use the bot’s response to improve your next request. Avoid specific jargon or slang.

Instead of “Create 10 social media posts for me,” say something else

for example

You are an expert in artificial intelligence. You want to give more keynote speeches. But your content is too technical and too deep. Create 10 theses on artificial intelligence with reference to the present, rate these theses on a scale from 1 (irrelevant) to 5 (relevant) and create hypotheses and suitable social media posts for LinkedIn. Present everything in a table. Work in simple language. LinkedIn posts should be 10 lines long and start and end with a question, contain emojis and 3 hasthags (at the end). Many thanks.

How is this prompt structured? Now in this case we have decided on a role-goal model, so we give the AI an indication of who we are and what we want, and also indicate the expectations of the result.

This is important, similar to a recipe, so that the AI can better understand which logic we are actually following. In this example we follow the structure of the ROLE – the GOAL – the INTENT – and then give CONTEXT (through SPECIAL or NOTES):

The result then looks like this:

You can then teach your AI something that it then uses, for example we always train our social media inputs on our 5:3:2 model and we then pass this on:

Based on this, the ideas become much more precise and easier to create and the results can then be modified:

Bonus tips when it comes to prompting

Avoid unnecessary filler words and vague wording in your queries. Be more specific and precise, e.g. B. “Write a 500-word essay on capitalism in the style of Allen Ginsberg.” Clear queries lead to more accurate results. Use the model’s knowledge and give it enough, but not too much, information. Understand the structure of AI prompts:

• The content type determines the focus, e.g. B. Drawing or sketch.
• The description includes details such as art style or resolution.
• The style defines the desired output style, such as Art Deco or Abstract.
• The composition describes the arrangement of the elements.

In theory, we can all benefit not only from prompting, but also from working with AI in general, and by preparing to work with these technologies we can maximize these benefits.

By paying attention to these components, you can optimize the results of your AI model. Quality of input determines quality of output.