Prompt Engineering for ChatGPT and Generative AI

There is no harder feeling than identifying the best tools to get something done and not being able to explore it efficiently. Over the years, there have been different Artificial Intelligence and Generative AI tools similar to ChatGPT but none of the free tools matched the complexity of ChatGPT chat features.

While the tool may be frowned upon by certain professionals, it is of no doubt highly beneficial and recommended for anyone who can use it effectively. These tools can be used to generate written content such as blog articles, emails, landing page copies and visuals for illustration purposes. In this article, I will share hacks on how great prompt engineering can help anyone get the best from Generative AI tools like ChatGPT, Midjourney and DALL-E.

ChatGPT was launched in November 2022 by OpenAI, leveraging OpenAI’s GPT-3.5 language models. The tool has gained viral attention from leading experts across the world and recorded over one million users in less than a month after the launch. There are several screenshots and testimonials from users on the internet on how helpful the tool has been from professionals to university students, bloggers and influencers.

Prompt engineering is a popular technical concept that explains the relationship between input (prompts) and outputs (results) for Generative AI models. Simply put, the quality and relevance of results that can be derived from Generative AI like ChatGPT are highly dependent on the quality and contextual mapping of the prompts or requests. Being able to craft great prompts can help get helpful responses from and ensure the NLP models understand the request and can equally provide relevant and accurate responses.

Here are the top five tips for writing quality request prompts for NLP models.

1. Build a conversation: This is of course the most important and highly untapped hack. Ensure you build conversations and provide enough context, just as you would when chatting with a human. Don’t just get into the prompts without building some background information to aid further decisions. Start by engaging the chatbot with some information about the field or request type, ask questions to validate assumptions, and then begin by asking questions in batches.

2. Keep it short and simple (KISS): Do away with complex sentence structure or Shakespearean vocabulary when typing request prompts. Make your prompts straightforward and as simple as possible. ChatGPT is designed to process and respond to simple requests, making it easier for you and the tool.

3. Specify requests: Ensure your prompt is specific and includes what ChatGPT should and not do. For example, instead of saying “Tell me how to distribute my article”, you can try “Tell me how to distribute my article online assuming I don’t have a Twitter account”. This will further help the model understand the objectives for the prompt and provide favourable responses.

4. Proper grammar and politeness: Engage with the bot politely and respectfully as it is designed to mimic human conversations. Ensure to use proper grammar and punctuation to help the model understands what you are asking.

5. Give rating and feedback: There is no better way to further enhance the quality of responses for ChatGPT than providing additional feedback to the team. Feel free to click on the thumbs icon (thumbs up preferably) and share some more information on what would the ideal answer have been.

   
   

Prompt engineering is being able to communicate clearly what a Generative AI should do and the only limit is the inability to express the exact briefs clearly. This can further help creatives, engineers, product managers and consultants generate great briefs to avoid wasting time and second-guessing. If done right, it can also help maintain consistent brand guidelines and get out-of-the-world inspiration or ideas.

I strongly believe following the above-listed tips and using specific, clear, and contextual request prompts, you can get the best responses from ChatGPT for a wide range of enquiries for your personal and professional endeavours.

Here are some examples of prompts and requests for different types of enquiries:

Engineering:

• “What is the formula for calculating the volume of a cylinder?”
• “Can you explain how a hydraulic system works?”
• “What is the principle behind the operation of a steam turbine?”

Medical enquiries:

• “What are the symptoms of the flu?”
• “How can I treat a sprained ankle at home?”
• “What are the risks of taking aspirin during pregnancy?”

Education:

• “Can you explain the process of photosynthesis?”
• “What is the significance of the Declaration of Independence?”
• “How do I solve this math problem: 2x + 3y = 12?”

Entertainment:

• “Tell me a joke about a penguin.”
• “Can you recommend a good book to read about drums?”
• “What are some interesting facts about the Great Wall of China?”

Conclusively, there will be more demands for “Prompt Engineers” in a few years given the rise of generative AI tools such as ChatGPT, Midjourney and DALL-E. In my next article, I will give in-depth examples of the terminologies for writing prompts using request mapping categories from fields/industries, formats, perspectives, limitations and boosters.