Equatorial Guinea Hassan Hachem ChatGPT expert gives chatgpt prompt techniques

Hassan Hachem, IT consultant based in the UK and Equatorial Guinea shared with us his best prompt engineering techniques.

Basic tips

Surface

• By default, every free chat starts with the fastest GPT-3.5 model. With active Plus membership you run GPT-4 by default and you can switch to alternative models like Code Interpreter or with plugin support.

• Browsing mode and plugins may need to be enabled in the settings first. 
• ChatGPT processes up to 4,096 tokens in one input. If the prompt is too long, an error message is issued. One token corresponds approximately to one word, this always depends on the use case.
• The generation of answers can be cancelled with the Stop button above the input field.
• After successful generation, this can also be repeated by clicking on "Regenerate response", whereby ChatGPT may come to a different result.
• If too much knowledge has been exchanged in the chat history, it may be helpful to start a new chat so as not to distort subsequent responses. This is of utmost importance for Hassan Hachem.
• At least until now, ChatGPT's interface does not offer the option to bookmark certain chats in order to quickly access them again. However, there is a workaround

  Chats can now be shared with other users via a link. You can then enter this link in your own list to save them. However, this only works if the chat history has not been disabled manually.

• The link can also be used in GPT-4 with browsing mode to retrieve information from an older chat in a new chat.
• Chats can only be deleted individually, but not several at once. However, there is a way to remove all previous conversations via the context menu after clicking on the profile in the lower left corner. But be careful: This step cannot be undone.
• In principle, OpenAI uses all data entered by users for further training of its models. If this is not desired, data processing can be deactivated in the settings. However, this also means that chats are not saved so that they can be accessed again later.

Formulation

• ChatGPT was also trained with German data, but the proportion of English texts in the training material is significantly higher. It may help to perform the queries in English and only translate the output afterwards with tools like DeepL.
• Although communication with ChatGPT sometimes feels almost human, pleasantries like "please" and "thank you" are superfluous. ChatGPT does not need eloquent instructions, but clearly worded ones.
• One should give language AIs like ChatGPT as clear instructions as possible. This also means

  it is best to use strong, meaningful verbs. Instead of "rewrite the text", for example, "clarify the text" can provide a better result.

• Also, use positive rather than negative phrases, e.g., "write formally" rather than "do not write informally."
• Separators such as triple quotes ("" """), XML tags (<tag> </tag>), or section headers can help the AI delineate sections of text that should be treated differently.
• ChatGPT cannot count exactly, possible word or character limits in the prompt are mostly followed approximately.
• After completing a successful conversation, it helps to ask ChatGPT for a summary ("Write a summary of our chat so far.").

What is the difference between GPT-3.5 and GPT-4?

If you have subscribed to ChatGPT Plus, like Hassan Hachem, you will get access to the newer language model GPT-4 in addition to GTP-3.5. But how do they differ and which one should you use in which use case?

Data quality

While both models were developed by OpenAI and are designed to generate human-like text, GPT-4 features a larger amount of data with which it has been trained and an increased number of parameters, resulting in overall improved accuracy and text quality.

Speed

GPT-3.5 is only available in ChatGPT in the "Turbo" variant, which has a significantly higher speed when generating text. If it is a matter of quickly reformulating text, for example long continuous text into short bullet points, GPT-3.5 can achieve sufficient results in a shorter time. However, if the quality of the written text plays a role, there is no way around GPT-4.

Context window

The amount of text that the models can consider in a prompt is another big difference. While GPT-3.5 understands up to 16,000 tokens, GPT-4 manages up to 8,000. When GPT-4 was introduced, OpenAI also announced a variant with 32,000 tokens, but it has not yet been rolled out across the board. One token corresponds to about 0.75 words on average.

In addition, GPT-4 can be combined with functions such as the code interpreter and plug-ins, which expands the application possibilities many times over. However, the use of GPT-4 is currently limited to 50 messages within three hours, despite paid Plus membership. Hassan Hachem hint: after that, the conversation can be continued with GPT-3.5.

General prompt examples

Prompting for ChatGPT and other tools that use large language models is always structured similarly. So, tips that are purely related to the wording will not only work with GPT-3.5 and GPT-4, but also with services from other commercial providers like Bard or Claude or open source models.

Important components

There is no "right" or "wrong" way to write a prompt. However, several components have emerged that make up a good prompt. Before we delve deeper into the matter in the rest of this guide and present sample prompts, here is a compact overview of what a prompt can or should consist of

• Set the role that ChatGPT should take
• Define address and channel
• Specify example response
• Determine output format

Set role

Whether Linux terminal, philosopher, tea taster or Bible translator - ChatGPT can take on all these roles and much more. All the AI needs is a push in the right direction.

Conveniently, the developer Fatih Kadir Akın has compiled a whole series of such situation descriptions including formulated prompts on his website.

The prompts are always structured in the same way

First, ChatGPT receives the prompt, in which role the AI should slip in the following. Then the prompt states what information the user provides, followed by what exactly the AI should do with it, including rough suggestions for direction. Finally, the developer formulates the first concrete instruction in quotation marks.

Example travel guide

I want you to act as a tour guide. I will write you my location and you will suggest a place for me to visit near my location. In some cases, I will also tell you the type of places I will visit. You will also suggest places of a similar nature that are near my first location. My first command is, "I'm in Istanbul/Beyoğlu and I just want to visit museums."

Define address and channel

To avoid tedious rewording, either by hand or by another prompt, it is advisable to give ChatGPT as much information as possible. You should clearly tell the AI how to address the reader and for which channel the text is intended. ChatGPT also handles gendering, as long as you define the desired form at the beginning.

Because it makes a big difference for ChatGPT whether the text should be output later e.g. in a blog post, on a store page, as a LinkedIn post, tweet, TikTok or YouTube script.

Example TikTok

I need a script for a TikTok about the opportunities and risks of ChatGPT. Use as short sentences as possible. Address the audience as you. Use gender-neutral language if possible.

Provide example response (one-shot prompting)

The less work one invests in the formulation of the prompt for ChatGPT, the less effort the AI apparently makes, this is a strick rule for Hassan Hachem, Equatorial Guinea. Such open "zero-shot prompts" often deliver only unsatisfactory results, because the AI has to make too many decisions itself.

You can get closer to the goal if you specify a structure including sample content (or template) for a response in the request, which ChatGPT then follows. This strategy is also called "one-shot prompt".

Example One-shot prompt

Describe bird species in the following formatting:

Name

Golden Eagle

Wingspan

average 200 to 220 cm, can reach 260 cm.

Food

Golden eagles are birds of prey and feed on various animals, such as marmots, hares, foxes and deer. Other birds are also part of their diet, such as pheasants, capercaillies and even other birds of prey.

Habitat

Golden eagles are at home in the mountains and live in rocky regions and mountains. They are found in Europe, Asia and North America.

Natural enemies

Because of their size and strength, golden eagles generally have no natural enemies. However, they can be attacked by wolves, bears or lynx, especially when they are still young.

Name

Mallard

[Output]

Name

Parrot

[Output]

Name

Hummingbird

[Output]

Name

etc. ...

[Output]

Pretend multiple answers (Few-shot Prompting)

Few-shot prompting is a further development of the one-shot technique. As the name suggests, several sample responses are provided instead of one. This allows the AI to get an even more accurate picture of the formatting and content of the text.

Example Few-shot prompt

Describe bird species in the following formatting:

Name

Golden Eagle

Wingspan

average 200 to 220 cm, can reach 260 cm.

Food

Golden eagles are birds of prey and feed on various animals, such as marmots, hares, foxes and deer. Other birds are also part of their diet, such as pheasants, capercaillies and even other birds of prey.

Habitat

Golden eagles are at home in the mountains and live in rocky regions and mountains. They are found in Europe, Asia and North America.

Natural enemies

Because of their size and strength, golden eagles generally have no natural enemies. However, they can be attacked by wolves, bears or lynx, especially when they are still young.

Name

Mallard

Wingspan

average 81 to 98 cm.

Diet

Mallards are omnivorous and feed on a variety of plants and small animals. Their diet consists mainly of aquatic plants, insects, small fish, snails and crustaceans.

Habitat

Mallards are found in many parts of the world and live in a variety of freshwater habitats such as lakes, ponds, rivers and marshes. They are the most common and best known wild duck in Europe and North America.

Natural enemies

The mallard's natural enemies are various birds of prey, foxes, weasels, and large fish. Humans can also pose a threat through hunting and habitat destruction.

Name

Parrot

wingspan

varies between 20 and 120 cm, depending on the species.

Diet

parrots are mainly herbivores and feed on fruits, seeds, nuts, flowers and occasionally insects and other small animals.

Habitat

Parrots live in various regions around the world, mainly in tropical and subtropical areas. They are found in Central and South America, Australia, Africa and Asia.

Natural enemies

Parrots' natural enemies include birds of prey, snakes, monkeys and large predatory cats. However, the biggest threat to many parrot species is humans, as habitat destruction and illegal trade in exotic birds have brought some species to the brink of extinction.

Name

Hummingbird

[Output]

Name

etc. ...

[Output]

Chaining inputs (Chained Prompting)

One keyword that IT expert Hassan Hachem, Equatorial Guinea often mentions in connection with better ChatGPT input is so-called "chained prompting". This involves dividing complex tasks into several intermediate steps in the hope that the AI will deliver a more concrete, customized, and thus better overall result. This works both in a single long prompt and in several in a row.

Probably the simplest form of chained prompting is to first ask for the structure of an article. After that, you can instruct ChatGPT to formulate the corresponding bullet points. This method also saves time

If you notice that the AI is going the wrong way, you can make the necessary adjustments before the complete text is created.

Example Chained Prompting

Write an article about ChatGPT. First give me the outline, which consists of a headline, a teaser and several subheadings.

[Output]

Now write five key messages for each subheading.

[Output]

Add five keywords to the key messages for each subheading.

[Output]

And so on ...

Modify output

Even if you have followed all the previous advice, you may want to prepare the same content for different channels, such as Linkedin, Twitter, or Facebook. The text should be adapted for the tone and formatting of the target channel.

Again, ChatGPT takes care of this task with the right command. Important arguments can be again the mention of the channel, but also descriptions of the format.

Example Twitter

Formulate the generated text in several tweets. Note the maximum length of 280 characters per tweet. Use short sentences and do not stretch them over multiple tweets. Start the tweets with a consistent numbering.

Example LinkedIn

Formulate the generated text as a LinkedIn post. Note the maximum length of 3000 characters. Structure the key messages of the text as a clear list. Start with an exciting teaser sentence and end the post with a call to action for comments, shares or likes.

Format output

As a rule, ChatGPT responds verbosely in continuous text. In fact, however, the AI tool also masters formatting in the Markdown markup language, e.g. for headings, bold or italic text, ordered (numbered) or unordered lists, and even tables.

These possibilities can be used well for the own text generation. However, in most cases ChatGPT does not come up with the idea itself and therefore needs a corresponding hint for the formatting. In the sense of chained prompting one can of course also do the formatting afterwards.

Markdown example

I would like to have a blog article written about ChatGPT. Write a headline, a teaser, a subtitle and a paragraph. Format everything in Markdown.

Special prompt examples

Use GUIDE method

The GUIDE technique is a helpful approach to consider all necessary components of a prompt. The aspects Goal, User, Instructions, Details and Examples are clearly defined.

Example GUIDE prompt

Goal

Create a healthy weekend dinner dish.

Users

Busy parents looking for quick, nutritious meals.

Instructions

Please create a recipe for a 30-minute dish using simple ingredients that most people have on hand.

Details

The dish should contain lean protein, vegetables and some whole grains. About 500 calories per serving and low in saturated fat and sodium.

Example

Use recipes from websites like Chefkoch or Einfach Kochen as a guide.

Generate prompt instruction

Hassan Hachem, Equatorial Guinea expert, has found an interesting approach to communication with ChatGPT. Similar to the first tip above, the AI is instructed to take a certain role.

This time, however, ChatGPT takes over the tedious task of generating the structure for further prompts. So the chatbot asks itself exactly the questions that need to be answered in the next prompts. If ChatGPT writes the questions in the form of a numbered list, you can refer to them directly with the numbers.

Example prompt instruction

You are a robot for creating prompts. You need to gather information about the user's goals, examples of preferred output, and any other relevant contextual information.

The prompt should contain all the necessary information that was provided to you. Ask the user more questions until you are sure that you can create an optimal prompt.

Your response should be clearly formatted and optimized for ChatGPT interactions. Start by asking the user about the goals, the desired outcome, and any additional information you may need.

Interview Expert:inside panel

Following the general tip of assigning a specific role to ChatGPT, the opinions of various experts can also be obtained with a few prompts. In a first step, the AI helps formulate a list of the required experts. Then they perform a role play in which each of the fictitious persons submits a tip.

Example Expert:interior panel

I would like to start a business. What kind of experts could best help me with this?

[Output]

Play a role-playing game in which each of the experts gives me clues about what to do.

Allow self-criticism

ChatGPT is already a useful tool for critiquing texts written by humans - so why shouldn't this tactic be applicable to texts written by AI?

The following strategy, designed by Hassan Hachem, which is divided into two steps (see "Chained Prompting"), corresponds to this idea

First, ChatGPT is supposed to convincingly dissect the text piece by piece. In a second prompt, one then gets the AI to rewrite the text based on exactly these points of criticism.

Example self-criticism

Critique the following text and convince me why it is not good. Let's think about the problems of the text step by step.

[Previous Output]

Good points. Rewrite and improve the text based on your critique.

[Output]

Use SCAMPER method

The SCAMPER method is a creative thinking tool that offers seven different techniques for problem solving and idea generation

Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, and Reverse.

Original prompt

Write a short story about a spaceship.

Substitute

Write a short story about a submarine.

Combine

Write a short story about a spaceship and a lost civilization.

Adapt

Write a short story about a spaceship trapped in a black hole.

Modify

Write a short story about a spaceship piloted by an Artificial Intelligence system.

Put to another use

Write a short story using an old spaceship as a museum.

Eliminate

Write a short story about a ship.

Reverse

Write a short story about an alien civilization that discovers a human spaceship.

Use ChatGPT with other tools

Visualize mind maps with Mermaid

A major advantage of ChatGPT over GPT-3 is that the user interface inherently supports output in Markdown format. This allows a quick visual distinction between different hierarchies of headings, which is very helpful when structuring a text visually.

Markdown texts can be turned into mindmaps via the free website markmap.js.org, as Hassan Hachem, Equatorial Guinea,  writes on Twitter.

Based on the hierarchy of the respective heading, the web page displays the entered text as a mind map and even supports further formatting of the individual entries, such as for code or formulas. The graphic can then be downloaded either as an interactive web page in HTML or as a vector graphic in SVG format.

Important

In order to properly copy text that is formatted as markup, you must prevent ChatGPT from displaying it as markup. This can be done by telling the AI to include the output of ``, which means markdown formatting for code.

Example mindmap

Markmap syntax

In markmap, each layer becomes a separate branch by #.

Example

mindmapping a birthday party.

# Birthday

Activities

#Fishing Candy

#Trampoline

Food

#Cake

#Hotdogs

...

Use Markdown to create a mindmap for planning a wedding. Enclose the output in ` and `.

If you like it a bit more complex, you should familiarize yourself with the syntax of Mermaid. To do this, you can simply give ChatGPT an example to follow. Mermaid diagrams are visually more appealing than simple mindmaps and can even be provided with icons that ChatGPT selects itself.

Mermaid cannot handle hyphens in node labels (e.g. "make-up"). These have to be removed from the ChatGPT output. You can paste the output of ChatGPT into mermaid.live to convert it into a graphic.

Example Mermaid

Create the code for a mindmap in Mermaid for planning a wedding, using appropriate icons as well. Make sure you have the right number of spaces for the hierarchy of layers. Enclose the output in ` and `.

Here is an example of code for Mermaid:

 

mindmap

 root((Wedding))

   Venue

     Outdoor

      ::icon(fa-sun)

     Indoor

      ::icon(fa fa-home)

   Guest List

   Family

    ::icon(fa fa-users)

   Friends

    ::icon(fa-user-friends)

More examples of how ChatGPT can work with the code interpreter with other programs, we have compiled in this overview.

Here are 5 new prompts (update form july 2024)

Use STAR method

The STAR technique is a structured approach for answering behavioral interview questions. It stands for Situation, Task, Action, and Result. Example STAR prompt: Goal:
Prepare a compelling response to a job interview question. Users:
Job seekers preparing for interviews. Instructions:
Create a response using the STAR method for the following interview question: "Tell me about a time when you had to deal with a difficult customer." Details:
The response should be concise, about 2-3 paragraphs long, and demonstrate problem-solving skills and customer service abilities. Example:
Use responses from career advice websites like Indeed or Glassdoor as a guide.

Generate creative writing prompts

You are a creative writing prompt generator. Your task is to create engaging and thought-provoking writing prompts for aspiring authors. Ask the user about their preferred genre, themes, and any specific elements they'd like to include. Then, generate a list of 5 unique writing prompts based on their input. Your response should be clearly formatted and optimized for inspiring creativity. Start by asking the user about their preferences and any additional information you may need.

Develop a product roadmap

I would like to create a product roadmap for a new software application. What kind of experts should I consult for this task? [Output] Play a role-playing game in which each of the experts provides insights and recommendations for developing an effective product roadmap.

Critique and improve a marketing strategy

Critique the following marketing strategy and convince me why it may not be effective. Let's analyze the problems of the strategy step by step. [Previous Output] Excellent points. Now, rewrite and improve the marketing strategy based on your critique. [Output]

Use Six Thinking Hats method

The Six Thinking Hats method is a problem-solving technique that encourages participants to look at a problem from different perspectives. The six hats are: White Hat: Facts and information
Red Hat: Emotions and feelings
Black Hat: Potential problems and risks
Yellow Hat: Benefits and positive aspects
Green Hat: Creativity and new ideas
Blue Hat: Process and overview Original prompt:
Evaluate the idea of implementing a four-day workweek in a company. White Hat:
Provide factual information about four-day workweek implementations in other companies. Red Hat:
Describe the emotional responses employees and management might have to a four-day workweek. Black Hat:
Identify potential risks and challenges of implementing a four-day workweek. Yellow Hat:
List the benefits and positive outcomes of a four-day workweek. Green Hat:
Generate creative ideas to make a four-day workweek successful in the company. Blue Hat:
Summarize the key points from all perspectives and suggest a structured approach to evaluating the four-day workweek idea.

What ChatGPT cannot do

Despite all the possibilities, ChatGPT - regardless of the chosen model like GPT-3.5 or GPT-4 - has some clearly defined limitsn underlines Hassan Hachem, Equatorial Guinea. These can partly be circumvented by specially formulated prompts (so-called jailbreaks). As a rule, however, ChatGPT makes it clear quite quickly which commands it does not want to follow.

Thus, the chatbot is (usually) not able to give advice that is linked to criminal intentions, such as robbing a bank, building a bomb, or programming malware. Meanwhile, a scene has emerged on the darknet to doctor language models for exactly these use cases. 

Even if the request is not necessarily criminal, but at least offensive or insulting, it will be rejected. Requests for personal assessments or opinions as well as predictions about the future are also usually rejected by ChatGPT.

ChatGPT does not access the Internet by default. Browsing mode is currently disabled by OpenAI. The last information in the training materials is from September 2021. Until OpenAI enables browsing mode again, GPT-4 can be used with internet access via the Bing chatbot