ChatGPT - How to Write Effective Prompts

Samy Boustany

ChatGPT

How to write effective prompts

​Introduction

Our ability to communicate effectively with the machine defines the limits of what we can achieve, in a world where ChatGPT, Claude, Mistral (the French model co-founded in April 2023 by Arthur Mensch, Guillaume Lample and Timothée Lacroix: the company raised 385 million euros in December 2023

and is valued at nearly 2 billion euros), Gemini and Copilot, etc., invade our news feeds, our tools, our JTs and our computers.

This short guide is designed for you, whether you're a professional or an individual, whatever your skill level or familiarity with generative AI (GenAI).

In just four chapters, you'll discover how to formulate prompts that reflect your expectations and guide AI towards producing relevant, personalized responses. My aim is to make prompting accessible to everyone by demystifying its fundamentals and equipping you with the tools you need to interact effectively and easily with AI. You'll be amazed at how clarity, specificity and intent can turn your prompts from simple requests into rich, fruitful conversations with GenAI. You'll learn how to frame your queries for specific results, how to use context and feedback to guide ChatGPT and all the other generative AI tools, and how to use their unique capabilities to enhance your work, creativity and learning.

It seems we need to get used to a world where humans and machines work together. By acquiring a better ability to communicate with linguists (a term proposed by Laurent Gouzènes, of the Fredrik R. Bull Institute, to refer to ChatGPT and company), we can ensure a smoother adaptation to this new revolution. What if prompting was child's play? My name is Samy Boustany, and I'm a consultant and

​innovation and marketing trainer. I'm co-founder of France's first prompt engineering course, and I'm supporting this transformation with companies and higher education schools, notably in collaboration with the FEDE

(European Federation of Schools). I began theorizing prompt methods in February 2023. I have trained thousands of people in initial and continuing training on the opportunities and limitations of these tools. I invite you to explore the fascinating world of prompting best practices through this short guide, designed to demystify and make this crucial skill accessible to everyone. If you have this book in hand, you're well on the way to understanding how to use generative AI tools effectively.

The era of generative artificial intelligence

You've already heard of ChatGPT, from OpenAI, founded in 2015 by current leader Sam Altam and Elon Musk (the latter having jumped ship in 2018 following management disputes). But do you know what the acronym GPT stands for? It stands for Generative Pre-trained Transformer.

— Generative means the ability to generate content of various kinds.

• Text: articles, scripts, dialogues.

• Images: logos, product designs, artwork.

• Music: compositions in various genres.

• Video: short videos, animations.

​• Code: programming for various languages.

• Voice: narration and text-to-speech.

— "Pre-trained: to function, a generative AI program needs to learn

things from lots (sometimes trillions) of data. This is known as training.

— Transform: this is a neural network architecture introduced by Vaswani and colleagues in 2017 in the article Attention is All You Need

(https://arxiv.org/pdf/1706.03762).

They have revolutionized natural language processing (NLP) and are the basis of the LLM (Large Language Model, ChatGPT being one of them), i.e. the ability to understand/generate so-called natural language (i.e.

as we humans use it) while calculating the most probable word by the AI.

This technology works in three stages:

1. Learning: the machine analyzes numerous texts to learn how words are usually associated.

2. Modeling: it builds a model that predicts the next word based on previous words.

3. Prediction: the AI uses its model to choose the most likely word to complete the sentence.

What is intelligence?

Here is the definition given by Larousse: "All the mental functions whose object is conceptual and rational knowledge (synonyms:

​conception, understanding, spirit, intellect, thought); ability of a human being to adapt to a situation, to choose means of action according to circumstances (synonyms: skill, dexterity, maestria, mastery, virtuosity)."

For some years now, when I've been speaking in public or giving training courses, I've liked to evoke the Swiss biologist, psychologist and epistemologist Jean Piaget's approach to intelligence: Intelligence is not what you know, but what you do when you don't know.

What is artificial intelligence?

Here's the definition proposed by the European Parliament: AI refers to the ability of a machine to reproduce human-related behaviors, such as reasoning, planning and creativity.

The term artificial intelligence was first used in 1956, at the Dartmouth Conference, by John McCarthy and his colleagues. Although criticized by some experts, it remains widely accepted today.

What is generative artificial intelligence?

ChatGPT, Midjourney, Claude, Mistral, Llama, etc., there are thousands of generative AI tools. They're also known as Large Language Models (LLMs).

At the end of this guide, you'll find a list of the tools I recommend and use every day.

According to the CNIL, generative artificial intelligence is a system capable of creating text, images or