How to prompt chatbots Generative AI

How to Prompt Chatbots (Generative AI)

Hello everyone, let’s get into part two of our series on Generative AI. In this episode, we’re going to discussing how to prompt Chatbots to get the best out of Generative AI (LLMs).

In our previous article, we introduced AI and generative AI. Generative AI isn’t the only type of AI. It is worth noting that there are other types of AI such as predictive AI which is used to analyze patterns in data and make accurate or predictive responses. This is very important in sectors like the stock market, transportation or even in predicting Natural disasters.

Artificial Intelligence Generative AI

Generative AI

Generative AI is a type of artificial intelligence that generates new content from existing content. In the last article, we highlighted that generative AI can be used to create images, compose music, and even generate text. Listed examples of these tools included Gemini, Copilot, Claude AI, and ChatGPT.

How do you get the best out of AI?

To get the best and most accurate information out of Chatbots like Copilot and Gemini, you will need to know how to prompt chatbots and how to engineer a prompt. There are rules and principles to observe in this process as well as levels of prompting which we will be highlighting underneath each prompt architecture. Prompting means generating a command that guides the bot toward a specific response.

How to prompt Chatbots: Prompt Architecture

AI tools like Copilot, and ChatGPT are examples of large Language Models (LLM). LLMs are software which are trained with billions of data, documents, journals, articles and sources from the internet. They can then use this information to generate content for you (i.e. Generative AI). In other for you to get the LLM to work correctly, you must know how to prompt it. By prompting it, you’ll tell it what it needs to do, how to do it, who or what it needs to be, and how it should respond to you. I will be academically giving examples.

1. Who:

This tells it what it needs to be. You are giving it a context on how to start responding to you. If you ask the AI or LLM to define a term or tell you about history, the response you may get is usually bland and generic e.g. give me a brief history of AI. You have to be more specific with your instructions, hence prompting. There are levels of prompting when focusing on “The WHO” e.g.

  • Simple: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as my History Teacher for a class discussion in history.
  • Intermediate: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as an expert Tutor in History for a class discussion in History
  • Complex: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as a historian and expert Tutor with high expertise in History for a class discussion on History.

2. What:

This tells the Chatbot what it needs to do. Remember that the “who” tells it who it needs to be, whilst the “what” tells it what it has to do. This will give you better answers because more context and detailed prompting helps it to deliver a higher quality response.

  • Simple: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as my History Teacher for a class discussion in history. I will ask you some history questions and please provide me with accurate answers.
  • Intermediate: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as my History Teacher for a class discussion in history. I will ask you some history questions and please provide me with accurate answers with sufficient details and ask me your questions.
  • Complex: Hello Copilot, I am a student currently studying History at a Nigerian University. I need you to act as my History Teacher for a class discussion in history. I will provide you with some history questions and please provide me with accurate answers with sufficient details, and examples and ask me your questions. I want your tone to be intriguing, engaging, and rational with a bit of humor when explaining some certain historical situation to a university undergraduate.

3. How:

This is how you want it to respond to your questions, with this you can tell it how it should provide answers, this could be in a step-by-step manner, or a bullet point. Furthermore, you can tell it to give you more details or provide external links to what it has given to you so you can expand your research.

  • Simple: “could you provide me with external sources for your response?”
  • Intermediate: I will need more details for your response, could you provide me with external sources for your response?
  • Complex: I will need more details for your response, please provide a clear outline of notable events in history in bullet points. Could you provide me with external sources for your response and links to them?

What you should know when prompting AI

1. Be Clear and Specific

Example: Instead of saying “Tell me about history,” you could say “Can you explain the causes of World War II?”

2. Provide Context

Example: “I’m writing a paper on climate change. Can you give me some recent statistics on global warming?”

3. Ask Open-Ended Questions

Example: “What are some creative ways to reduce plastic waste in everyday life?”

4. Use Keywords

Example: “Explain the concept of photosynthesis in plants.”

5. Be Polite and Friendly

Example: “Could you please help me understand how to solve this math problem?”

6. Specify the Format

Example: “Can you summarize this article in three bullet points?”

Limitations of Chatbots (LLMs)

However, it would help if you understood that there are some limitations to these chatbots. Prime examples are they may give you outdated responses or you would have to subscribe to a premium plan to use a better AI engine. For example, the free version of Chat GPT using GPT 3.5 was last updated in 2021 and GPT 4.0 (and above) is only available as a paid subscription service.

For you to use GPT 4.0 for free I recommend you open a Microsoft Account, sign into Bing or Microsoft Copilot, and set it to Creative mode which is more useful than using regular ChatGPT especially if you can’t afford the premium version.

Whatever the reasons you have for using AI, you will come to appreciate the fact that AI can be an important companion for information and learning new kinds of things. Another fact that you should take into consideration when using this tool is that it is prone to error and mistakes and you have to use other sources to verify your information. Although AI is designed to mimic human intelligence and even surpass it, it still needs a human in its development which is explained in Human in the Loop Model.


Please leave a comment if you have any difficulty and remember to:

4 comments

  1. Brilliant post…. Jeffrey Ogodogun you are just the best… AI is prone to errors at times… I noticed it in my Meta AI on WhatsApp… It failed a complex Physics question I had given it to solve to test it’s capabilities… So AI is limited

Leave a Reply

Your email address will not be published. Required fields are marked *