Becoming Efficient with ChatGPT. Tips for Prompting and Keeping Authenticity

Release date: 7 Dec 2024

Views: 86

Preface

These notes are for a presentation I gave at my local community event for office bearers.

The topic of the presentation was the use of A.I and how it can help us become more productive. I showcased what you can currently do with ChatGPT because it's the most popular and accessible for everyone.

I share:

Introduction

I have been using ChatGPT pretty much since it came out 2 years ago. I remember in the beginning you had to fully sign up and verify your phone number to be able to use GPT.

But just recently I found out that you can use GPT without even having an account. Just go on chatgpt.com and start prompting.

I have also used GPT API's to build web applications that involve AI work.

Aside from GPT I have also tried Claude, Gemini and even some of the LLM models released by meta. It's very clear that this tech is useful and it's here to stay.

Plus it's becoming more accessible day by day. Like I said earlier you can use a lot of these models these day completely for free. That is only going to get better with time.

As it goes for any tool, if it's part of your daily life you should try to optimise how you use it and become more proficient with it.

In this article I share the stuff that I have learned over the years so it can help you hopefully use the tool more efficiently.

For most information work I've split prompting to ChatGPT in two categories. I'm not saying everything fits into these categories but my aim is to help as many people as possible.

These two categories are:

  1. Expanding
  2. Condensing

I will showcase how to efficiently prompt GPT that apply to both of these categories and also share template prompts.

Expanding

What does expanding mean?

Let's say you have a report to write. You are not sure what the report as a whole looks like. But you do know what the contents of the report will be.

I call this expanding content.

You have an idea and you want to utilise ChatGPT to see transform that vision in your head to actual text on paper (or a file).

Instead of just asking GPT for your desired output (e.g, report, essay, letter) in a single sentence. Try to give it as much information as possible.

This information includes the targeted audience, the format, the tone, your key points and anything else that you think is relevant to generate your desired output.

Doing this will give you a much better head start than a single sentence.

The problem with writing a single sentence to try and get your desired output is you start to become frustrated.

The constant back and forth for 10+ prompts where you are trying to explain GPT what you actually mean makes you annoyed and feel like the tool is not good and it's not understanding you.

What I find useful is saving generic prompts that you can customise and add more details to for each different task / project you have.

The following prompt is a good one to start with and you may use it for your future work.

I have also added bullet points to explain what each section means. So you can clearly understand what the purpose of each section is.

Generic prompt to help you expand content

I have an idea for [TOPIC/PROJECT] that I’d like to expand into detailed content. Please help me create [FORMAT] tailored for [AUDIENCE]. Here's the initial idea or outline:
[OUTLINE/IDEA]

Expand this into a cohesive and detailed [CONTENT TYPE], ensuring it is:

For example, include:

  1. [EXAMPLES/EXTRA DETAILS]

Comments on brackets:

Condensing

This is when you are trying to do research work. You are going through lots of documents and websites to collate information.

Your plan is to consume that content and based on that write your own conclusions / summaries.

This process of collating is what I call condensing content. Going from a large piece of information into a tailored summary that applies to your needs and audience.

The same principles of being more specific apply to this category as well so it's not worth repeating. But try as much as possible to give GPT all the information you have that will help you get your desired output.

Generic prompt to help you condense content

I’m working through [SOURCE MATERIALS] and need to condense them into a focused summary for [PURPOSE]. Please help me create a [FORMAT] for [AUDIENCE], based on the following content:
[SOURCE MATERIAL/CONTENT TO SUMMARISE]

The summary should:

  1. Highlight key points such as [KEY INSIGHTS]
  2. Eliminate unnecessary details or repetition
  3. Adapt the tone and structure to [AUDIENCE STYLE OR CONTEXT]

Comments on brackets:

My Biggest Tip to Use GPT

The one thing that I don't see get mentioned a lot is the big session memory capacity of each conversation. Realising that can change the way you structure your prompts and use GPT.

Let me explain.

When I say session memory capacity what do I mean?

Each time you start a new instance of a chat with ChatGPT, it has a set capacity of how long the conversation can go. If you exchange a lot of messages in the same thread you will eventually hit a limit and will no longer be able to message in that thread.

That means you've reached the limit.

Fortunately, the memory capacity of a given chat has increased a lot compared to before. Meaning you can actually discuss a topic for a long period of time before we run out of memory.

We can use this to our advantage.

Instead of trying to get all the required work done in one single message exchange, we can actually slowly build up to the final required output. This results in an overall better quality of the final output produced.

Because you are able to drive GPT in a much more structured way.

Let's use an example.

Task: Generating a monthly report of community's activities.

The one shot message method

Prompt: Generate a report about this month's activities for my community. We played football, held 2 Friday classes and also went out for a meal.

With this approach you will get a report by GPT but you don't have much control over the response.

You will find yourself writing a lot of further messages out of annoyance because it didn't understand what you meant the first time or you missed out something.

This builds frustration and requires a lot more time to make micro changes to get your desired response.

Overall this makes working with GPT dissatisfying.

Generating the report in an iterative way

1st Message: I want to write the monthly report for my community's activities. Let's start by building a structure first and then we will add content later. I want to have 3 different sections. Physical health, religious education and community bonding. Can you build a structure for this report?

Notice how you have much more control. We create this prompt knowing that what we will get right after won't be the final answer. It will just be the structure.

Once we have the structure of the report as we desired we can ask GPT to add the content into those sections.

This way we have more control and we aren't frustrated by not getting the desired answer in the first go.

Prompting like this is only possible due to the increased memory capacity.

Where I Think You Shouldn't Use GPT

Communicating with others, especially messaging. That is the one thing where I think we shouldn't use A.I.

Here's why.

If you are a part of a community, chances are when you message people you aim to build a connection with them. Especially if you hold a position where you are the person responsible for organising events for your community.

In matters like these, authenticity is THE most important thing.

If the person on the other side realises you aren't REALLY writing these messages then it contributes towards moving that person away from your community.

This is more damaging than if this happens to you with a 1:1 interaction that isn't related to your community.

Because you aren't just damaging a connection between you and the receiver, you're negatively impacting the community as a whole due to a person moving further away from it.

A key thing to realise is, the more A.I you use in a written piece of text the less of your own authentic voice you have. Excessive usage makes you end up sounding like the average of everything that is already available on the internet.

If you don't have your own authentic voice when talking to others you will struggle to build meaningful connections.

Those are my reasons as to why you maybe should avoid use of A.I when communicating with others in your personal circles.

I hope you found this read useful :)

Join Snapshot weekly letter by maghfoor. 📬

My goal with the newsletter is to provide as much useful information as possible that will help YOU in your life. Topics can include engineering, business, career, and life.

No spam. Ever. Unsubscribe anytime.