June 2, 2023 - Business The art of prompts with ChatGpt
Updated on 13th of July, 2023
In this article, I discuss the art of the prompt on ChatGpt, and give you a few tips on how to use it, as well as sharing my opinion.
Art of Prompt
ChatGPT, developed by OpenAI, is an artificial intelligence that will revolutionize many people’s everyday working lives. It is capable of interacting with users in natural language. However, to maximize its potential, it’s essential to understand the art of formulating effective prompts (directions or questions). In this article, we guide you through the nuances of interacting with ChatGPT, what you can get out of it, and provide examples of successful exchanges.
The idea is to show you how this tool can become a precious ally in your daily life.
1. Clarifying the question
One of the keys to getting useful answers from ChatGPT is to ask clear, precise questions. The precision of the request influences the quality and relevance of the answer. For example, asking “What is photosynthesis?” will generate a different response than asking “Can you explain photosynthesis in simple terms for an 8-year-old?”. Precision allows you to tailor the AI response to your specific needs.
2. Context ans details
Providing context can help ChatGPT give more precise and relevant answers. For example, if you’re asking for help writing a cover letter, mention the sector or position you’re applying for. A letter for a job in finance won’t sound like a letter for a position as an English teacher.
Dialogue can take place over several exchanges. It may be useful, for example, to explain your persona, your role, the name of your company, and so on.
It is also possible to give instructions and check that they have been taken into account by asking “Have you understood?
It is also possible to influence the result by giving instructions, such as the number of output words, or the desired level of synthesis or development.
In use, I have to admit, it’s absolutely stunning.
3. Experimentation
Feel free to experiment and iterate with different formulations to see how the AI reacts. ChatGPT is a robust AI capable of understanding a variety of prompts, but it is sometimes necessary to repeat or rephrase a question to get the desired result.
Precise instructions and specific requirements
Sometimes it’s necessary to ask ChatGPT to respond in a specific way. Whether it’s to obtain verified information or to request a specific presentation of results, AI is flexible enough to adapt to a variety of requests.
Examples of effective exchanges
Example 1: For writing content.
Prompt : “ChatGPT, can you write a brief summary of a scientific article on AI in healthcare?”
Response: “Artificial intelligence (AI) is increasingly used in healthcare to improve diagnosis and treatment. The article explores several AI applications, including identifying diseases from medical images and predicting disease progression from real-time health data. It concludes that AI has the potential to transform healthcare, although some challenges need to be overcome, such as data privacy issues and public acceptance.”
Example 2: For learning
Prompt: “ChatGPT, can you explain Schrödinger’s equation in simple terms?”
Answer: “The Schrödinger equation is a central formula in quantum physics that describes how the quantum state of a physical system changes with time. Imagine you have a wave on a pond. This wave can be big or small, fast or slow. Schrödinger’s equation is like a recipe that tells you how that wave evolves and transforms over time.”
Example 3: For assistance with everyday tasks.
Prompt: “ChatGPT, can you help me create a shopping list for a vegetarian dinner for four people?”
Response: “Sure, here’s a basic shopping list :
- Quinoa: 1 cup
- Mixed vegetables (peppers, broccoli, carrots): approx. 1 kilo
- Goat’s cheese: 200 grams
- Olive oil, salt and pepper
- Garlic: 2 cloves
- Fresh basil
- Baguette bread: 1
- Dessert: fresh fruit or sorbet”.
Exemple 4 : Pour la recherche d’informations vérifiées et formatées
Prompt: “ChatGPT, can you provide a bulleted list of verified facts about the environmental impact of meat production?”Prompt: “ChatGPT, can you provide a bulleted list of verified facts about the environmental impact of meat production?”
Response: “Of course, here are some facts :
- Meat production is one of the main causes of climate change, accounting for around 14.5% of all man-made greenhouse gas emissions (source: FAO).
- It contributes to deforestation, particularly in Amazonia, where forests are often cleared to make way for pastureland.
- It is also a major source of water pollution due to the run-off of fertilizers and animal waste.”
Example 5: Assistance with code generation
Prompt: “ChatGPT, can you provide me with some simple HTML code for a web page that contains a header, a paragraph of text, and an image?”
And here’s an example of how ChatGPT could answer that prompt:
Response
In this example, you should replace “url_of_your_image.jpg” with the actual URL of the image you want to display, and “Image description” with a text description of the image.
It’s important to understand that ChatGPT can generate sample code, but cannot write complex or project-specific code. For such tasks, it’s best to consult a developer or other programming expert.
A powerful tool
ChatGPT is a powerful tool that, when used effectively, can transform the way you work, learn and communicate. The art of effective prompts relies on precision, context, precise instructions and experimentation. With these tips in hand, you’re ready to interact with ChatGPT to the full.
But it has its limits.
It’s important to bear in mind the limitations of this tool, and to put it in its rightful place. It’s so impressive and bluffing that it could almost make us lose common sense 😉
| ChatGPT Limitations | Description |
|---|---|
| Knowledge break (1) | ChatGPT has a knowledge break, i.e. a date up to which it has been trained on data. For the GPT-3 version, this date is September 2021. Consequently, it cannot provide information on events or developments occurring after this date. |
| Lack of understanding | ChatGPT doesn’t understand the meaning of information the way a human does. It generates responses based on patterns it has learned through training, not on a true understanding of the content. |
| Lack of reliable information | Occasionally, ChatGPT may produce inaccurate or misleading information. It does not have the capacity to verify the accuracy of its answers or to consult up-to-date sources. |
| Difficulty with complex or nuanced questions | If a question requires expert knowledge or nuanced understanding, ChatGPT may not be able to provide an accurate and appropriate answer. |
| Tendency to develop answers | Sometimes, even when ChatGPT isn’t sure of an answer, it may try to “guess” or extrapolate, which can lead to incorrect information. |
| Lack of long-term context | ChatGPT is limited in its ability to remember the context of long conversations. It is optimized to respond to immediately preceding messages, but can lose track of longer discussions. |
| Ethical and safety limits | ChatGPT is designed to refuse to generate certain types of content, such as explicitly illegal content, violent content or content that could be used to deceive or manipulate. However, it can sometimes fail to respect these restrictions. |
(1) : you can partially overcome this problem by implementing the chrome plugin webchatgpt.
Example: you want to know the glorious story of Count duschmoll, who lived in the 16th century and conquered all of Europe? no problem ChatGpt will invent your fantasy.
Here again, observe the finesse and nuance of ChatGpt.
If you ask him about the history of the duschmoll account, he’ll tell you that the information is unverified and that he can’t give you an answer. But if you ask him to tell the story, he’ll gladly do so.
By the way, I can’t resist sharing with you this reply from ChatGpt on this point:
In the 16th century, an unforgettable figure appeared on the European scene, making kings and queens tremble with his audacity and ingenuity. His name? Count Duschmoll.
Count Duschmoll was born into the nobility of the small principality of Liechtenstein in 1500. Despite his humble origins, the young Duschmoll showed great intelligence and ambition. These qualities attracted the attention of the court, which took him under its wing and provided him with a first-rate education.
In 1520, Duschmoll left his native Liechtenstein to embark on a journey across Europe. Armed with his shrewdness and charisma, he toured the European courts, forging alliances, negotiating treaties and making influential friends.
In 1525, Duschmoll returned to Liechtenstein and assembled an army with the help of his many allies. The following year, he launched a campaign of conquest that overturned the established order in Europe. Duschmoll and his army crossed the Alps, conquered northern Italy, then marched westwards, crossing France and reaching the Atlantic in 1527.
Duschmoll didn’t stop there. In 1529, he turned his attention northwards, marching on England and Scotland before crossing the North Sea to take control of the Nordic countries. In 1530, he reached Eastern Europe, adding the vast territories of Poland and Russia to his empire.
In just ten years, Duschmoll had accomplished what no one had ever done before: he had united the whole of Europe under a single banner. His reputation as a fair and enlightened ruler won him the loyalty of his subjects, and enabled him to maintain peace and stability in his vast empire.
Count Duschmoll reigned until his death in 1570, and his legacy continued well beyond his reign, shaping European history for centuries to come. Even today, his name is synonymous with vision, courage and leadership. His glorious story, though fictional, continues to fascinate and inspire.
So is chatGpt a monster that will discard humans?
ChatGpt isn’t really a virtual advisor, savant or writer. In fact, I asked the question, let him define himself:
| What chatGPT is | What chatGPT is Not |
|---|---|
| An AI Language Model | A Human |
| A Text Generator | A specialized consultant |
| A tool to generate answers based on the information it has been trained on | A software developer |
| A tool for idea generation and inspiration | A scientific expert |
| A tool for general explanations on a wide variety of subjects | A scholar with expert knowledge in specific fields |
| Able to generate creative content, such as stories or poems | A professional writer |
| Useful for practicing foreign languages | A professional foreign language tutor |
This is a general description, but it is important to note that ChatGPT has limitations and does not replace professional advice in specific areas. It is designed to help generate ideas and inspiration, and can give general explanations on a wide variety of subjects, but it is no substitute for human expertise.
It’s fair to ask whether this answer is entirely complete, given ChatGpt’s power as a tool for accessing knowledge. It’s quite clear to me that access to information is becoming easier all the time, and knowledge workers have a lot to worry about.
However, it’s clear that this tool, while enabling us to get help, doesn’t really reason and isn’t capable of moving the world forward. It does, however, provide access to information, and what used to be the privilege of an elite is becoming increasingly accessible. But it is not yet able to decide for us, to assess situations fairly or to change a situation.
Will ChatGpt make humanity better ?
ChatGPT, and artificial intelligence in general, have the potential to bring many benefits to society, such as access to information, assistance with learning, simplification of complex tasks and much more. AI can help democratize access to information and education, make processes more efficient and solve complex problems.
However, with regard to the idea that ChatGPT could “force humanity to think better before acting”, this is a trickier point. AI tools, like ChatGPT, can help provide information and inform decision-making, but they are no substitute for human judgment or ethical and critical reflection.
Ultimately, AI’s potential to make humanity “better” depends less on the technology itself than on how we, as a society, choose to use it. This means careful regulation, ethical and responsible use of technology, and constant awareness of AI’s benefits and limits.
Conclusion
Please note that a significant part of this article was written with the assistance of ChatGpt. But with the right prompts!
I have to admit that I was as surprised by ChatGpt as many others: I’ve taken courses in machine learning and deep learning, and I observed results in ChatGpt that I thought were unattainable in the state of the art. It’s very disturbing to think that such results can be obtained with algorithms of this type alone.
This obliges us to a certain humility as living beings endowed with consciousness…
I think that ChatGpt’s requirement is for future generations to be able to think, decide, find new paths and move towards the progress of knowledge. So the misuse of ChatGpt is to try to make it do the job for us. Because it doesn’t really think, it just renders.
But to think, you need a knowledge base that allows a certain degree of openness.
Education must therefore take this new situation into account: information is easily accessible, but it’s important to continue learning so that you can think and interact correctly with your virtual assistant to make the right decisions.
I don’t know if the national education system will recover.
It is in the interest of the younger generations to raise their level of awareness and adapt intelligently to this new situation. Society needs to reflect on the value of work, working conditions and the sharing of the world’s finite resources.
We’re living in a time of mind-boggling digital revolution. We’re at a crossroads. Fascinating.
But perhaps I’m getting off track now. Thank you for reading this article and I hope it gives you some useful information.
The author of this article, Stéphane Hosemans, is a CIO consultant. He has participated in numerous strategic projects involving information systems. In the frame of his Strategy and Digital Transformation activities, finding reliable information quickly and efficiently is a key activity for successful missions.
