Writing a book with ChatGPT

The recent arrival of chat-generated Artificial Intelligence, in the form of ChatGPT, prompted me to undertake a task I would have otherwise thought far too daunting – creating a reader-friendly version of the Bible.

My wife Lisa often told me that I should write such a book, for an audience that found the original Bible too difficult to read, and who didn’t necessarily understand the geography, politics and other aspects of the Bible stories.

The end result has just been self-published on Amazon Kindle as “The Good Book: An atheist’s guide to the Bible, its context, and its impacts“.

This short post is not about the book itself, but how I went about using ChatGPT to develop it.

Firstly, I asked ChatGPT to list me all the books of the Old and New Testaments, to establish the overall structure before tackling the contents. I also included an overview section at the start of the Old and the New Testaments, leaving some placeholders sub-headings to be determined later as I explored the content. Eventually, I also decided to add a discussion section at the end about the Bible’s positive and negative impacts on society, using a similar place-holder approach.

Then, I systematically asked ChatGPT about each of the 66 books of the Bible, and used the answers to start populating the content of the book. For each section, I then applied certain questions of interest, for example “What are some of the best-known stories in the book of XXX”, “Who are the main characters in the book of YYY and how are they related”, or “Describe the cultural and political context around the time of ZZZ”.

Slowly and steadily, the content of the document was built up – initially as mostly dot-points and repetitious-sounding chatbot language. Through several readings, and several rounds of Lisa’s scrupulous editing and feedback, the content eventually blended into a readable flow.

As the book progressed, my chat questions became increasingly more specific, for example “is halal similar to kosher”, “how many James’s are there in the Bible”, and “did Jesus walk on water before or after he calmed the storm”. Eventually I reached a point where I didn’t feel any more information would add value, and from that point it was just a case of following the online Kindle publishing tips.

My only warning about ChatGPT is that when using it, you need to keep in mind its key limitations:

• May occasionally generate incorrect information;
• May occasionally produce harmful instructions or biased content; and
• Limited knowledge of world and events after 2021.

An example of the first point above is that if you ask ChatGPT “Who is James in the Bible?”, it will correctly name and describe the three James’s (sometimes referred to as James the Greater, James the Less, and James the Just); but if you ask “Are there two James’s in the Bible” it may incorrectly agree “Yes, there are two James mentioned…”; or if you ask “How many James’s are there in the Bible” it may say “In the Bible, there are four individuals named James who are mentioned”, with an obvious double-up.

The lesson in relation to the questioning is therefore to:

• Carefully phrase your questions;
• Double-check the purported facts by asking similar questions in different ways;
• Make use of ChatGPT’s ability to refer back to earlier questions and answers within each session.

I’ll leave the final word in this blog post to ChatGPT…

Question – “What is the best advice you could give someone who has never used ChatGPT?”

ChatGPT – “The best advice I can give to someone who has never used ChatGPT before is as follows:

• Be clear and specific: When asking a question or providing an instruction, try to be as clear and specific as possible. This helps the model understand your intent accurately and provide a more relevant response.

• Break down complex questions: If you have a complex question or multiple sub-questions, consider breaking them down into simpler parts. This can improve the model’s ability to address each component effectively.

• Provide context: If your question or request requires some background information or context, it can be helpful to provide that information upfront. Giving relevant details can enhance the accuracy and relevance of the responses you receive.

• Be patient and iterate: Sometimes the initial response may not fully address your question or need. In such cases, feel free to ask for clarification or request additional information. Iterating the conversation can lead to better results.

• Use system instructions: You can provide high-level instructions to ChatGPT at the beginning of the conversation to guide its behavior. For example, you can specify that you want a detailed explanation or a creative story. This helps set the tone for the conversation.

• Verify information: While ChatGPT strives to provide accurate and helpful information, it’s always a good practice to verify important or sensitive information from reliable sources.

Remember, ChatGPT is an AI language model and may have limitations and biases. It’s important to use critical thinking and evaluate the responses accordingly.”