Basics, Best Practices, Examples
What are AI prompts?
What are examples of available models online where I can enter AI prompts?
Why do we write AI prompts?
- In simple terms, an AI prompt is a way of communicating with an AI model by giving it some input text that tells it what to do or what to generate.
- We write AI prompts to tell the AI model what we want it to do or what we want it to create. An AI prompt gives the AI model some guidance and direction for generating the output. Without a prompt, the AI model might not know what to do or what to create.
- For example, if we want the AI model to provide the market size of the global automotive industry, we can write a simple prompt like “provide the market size of the global automotive industry”. Here’s an example of using this prompt in Bing Chat:
How are AI prompts different from normal questions?
- AI prompts are specifically designed to elicit a certain type of response from an AI model.
- They are often structured in a way that helps the AI understand the context and intent of the user’s message.
- On the other hand, normal questions can be more open-ended and may not always provide enough information for an AI to generate a relevant response.
- An example of an AI prompt could be “Find the top 5 articles on the benefits of meditation.”
- This prompt provides clear instructions for the AI to perform a web search and return a list of relevant articles on the topic.
- A normal question on the same topic could be “What are some benefits of meditation?”
- This question may not provide enough information for an AI to generate a relevant response without performing additional research.
Best practices for writing AI prompts
Be specific and clear in your instructions.
Provide enough context for the AI to understand the intent of the prompt.
Use language that is easily understood by the AI.
For example: “Summarize the main points of the article ‘The Benefits of Exercise for Mental Health’ by John Doe in 3–5 sentences.” This prompt is specific and clear in its instructions, including the length of the response, and uses language that is easily understood by the AI.
Test and refine your prompts to ensure they elicit the desired response from the AI.
To test and refine AI prompts, you can try using them with an AI model and evaluating the responses generated.
If the responses are not what you expected or desired, you can revise the prompt to provide more specific instructions or additional context. (see below)
Here is an example of testing and refining your prompts based on the first response:
Context: I want to generate a 2-3 sentence summary of each of the major accomplishments of Jimmy Carter.
At first, the AI model misunderstood me and provided an overall 2-3 sentence summary, instead of describing each accomplishment in 2–3 sentences.
So, I refined the prompt based on the initial response in the following way to then get the desired results:
Common Challenges/Areas to Troubleshoot
Ensure that the AI understands the context and intent of the search query is important. For example, if a user is searching for information on “apple,” the AI must be able to determine whether the user is looking for information on the fruit or the technology company.
Ensure that the AI model presents accurate and reliable information.AI does not always provide the most accurate and reliable information and use the most credible sources.
To overcome this challenge, we can prompt the AI model to include and exclude certain types of sources. Here’s an example: