👩‍🔬Prompt Engineering: The Art of Crafting Effective Prompts

👋 Hello, and welcome to this new edition!

Today, we will cover ‘Prompt Engineering’ and ‘Prompting’.

You will learn:

• 👩‍💻The difference between technical and non-technical prompt engineering

• 📉 Figure out whether this is a short-lived trend or it’s here to stay

• 📚 Expand your arsenal of resources (courses, books, tools, communities) to master the art

🙌 Let’s dive in!

—

With the rise of the AI craze, there’s been a peak of interest in ‘prompt engineering’ from Google Trends, peaking at the beginning of this year, but still keeping a high volume after that (fun fact: the #1 country leading this trend is China).

‘Technical’ Prompt Engineering

Before we go further, I thought it would be useful just to make a note: there’s a technical prompt engineering, and then there’s a non technical one. The former is a very technical kind of role, which requires software engineering skills, that will shape and improve the software capabilities to respond and 'reason' when interacting with humans:

• 💡 Prompt engineering focuses on developing and optimizing prompts for efficient use of large language models (LLMs) in various applications and research topics.

• 💡 Prompt engineering skills help in understanding the capabilities and limitations of LLMs and are valuable for tasks like question answering and arithmetic reasoning.

For those interested in the details and in learning, you can check a really good guide here.

What about non-technical Prompting?

I think this is better suited to be called ‘prompting’:

• 🔧 Non-technical prompt engineering involves refining prompts for AI systems without requiring technical expertise

• 🎯 It focuses on optimizing the human input to generate desired outputs

• 📝 Non-technical prompt engineering considers factors such as clarity, specificity, context, and desired outcomes

The first thought we have when thinking about AI is that it automatically brings efficiency into work. That's not true. The more time passes by, the more we realize a few things:

Garbage-in, Garbage-out: This is a famous saying in Computer Science, simply meaning that if the data quality or input into software is not good, you will get a bad result from a machine too. In Prompting, this means that if you give AI a bad prompt, you will get a bad result.

Authenticity - Everyone can use AI, but if everyone uses it the same way, we're just going to get a bunch of copy-cat results. How can you keep authentic, human-like, and valuable content? By improving and customizing Prompting.

Productivity - All of the time you interact with an AI Chatbot is now time you need to optimize. How many iterations to get the result you needed? Improving Prompting is key also here.

Currently, we've been handed a new toy, but don't know how to play with it. I think Prompting is again like learning a new language; it's similarly a way to code, but with words.

Of course, as time goes by, we will see improvements in how the software will be able to interpret the human input, and we will need to do less work to optimize it.

There’s currently two main schools of thought:

Prompting will vanish in a year - Of course, as time goes by, we will see improvements in how the software will be able to interpret the human input, and we will need to do less work to optimize it. Also, tech companies are going to make more ‘no-prompt’ user interfaces, so that based on specific use cases, we won’t need to feed complicated instructions, but push buttons and give simple prompts.

Prompting will be here to stay - Even if user interfaces improve, you will always need to use prompting, and possibly with the evolution of AI technology itself, we will have even more conversations with machines. Maybe more colloquial and natural, but hey - don’t we have communication issues even with our fellow human beings too?

Expanding your Prompting Arsenal

To help you improve and master the art of prompting, here are several resources worth exploring:

Communities:

• Join Facebook groups such as "mj.prompt.tricks" focused specifically on MidJourney.

• Engage with the "WritingPrompts" subreddit, a vibrant community with 16 million subscribers.

• Discover insights and connect with like-minded individuals on Discord, ‘Learn Prompting’

Prompt Marketplaces

Embrace the "Gold Rush" of prompting by exploring marketplaces where prompts can be bought and sold. These platforms offer inspiration and quick solutions to prompt-related challenges.

• PromptBase: An online market where you can purchase and sell high-quality prompts that deliver superior outcomes, reduce API expenses, and allow you to market your own prompts.

• PromptHero: The largest prompt library for AI picture generating models like Stable Diffusion, DALL-E, Midjourney, and others. It offers millions of photos produced by artificial intelligence using various advanced AI models.

• Arthub.ai: A creative community dedicated to exhibiting, discovering, and creating AI-generated art. It provides access to thousands of AI-generated artwork and photos.

• PromptSea: A platform aiming to remove barriers and complexities in AI-generated content creation, offering creative freedom and equal opportunities for individuals and small studios. It allows easy and affordable creation of high-quality content with the assistance of AI.

• Visualise.ai: A marketplace for buying and selling prompts for DALLE, GPT-3, and Midjourney. It provides a platform to purchase text statements (prompts) that are fed into complex AI models to produce specific results, such as generating images of specific objects in specific styles.

Libraries and Books

Expand your repertoire of prompts by exploring libraries and books dedicated to the craft. These resources provide a wealth of generic and job-specific prompts to ignite your creativity.

• ChatGPT Prompt Book: Consists of over 300 unique writing prompts generated by the ChatGPT language model, designed for creative thinking and finding new ideas and perspectives.

• OpenAI Best Practices: OpenAI's GitHub repository consisting of best practices for prompt engineering, including the OpenAI Prompt Cookbook.

• PromptPapers: A GitHub repository containing a collection of essential papers on tuning pre-trained language models using prompts and other useful learning materials.

• DALL-E Prompt Book: A downloadable guide in PDF format that includes fundamental tips for prompts related to photography, illustration, 3D styles, and other topics, to leverage the complete potential of DALL-E.

• Awesome ChatGPT Prompts: A repository of diverse prompts that can be used with ChatGPT, encouraging users to expand the list with their own prompts and create new ones.

Interactive Tools

These take a more hands-on approach with interactive prompting tools that can enhance your prompt optimization efforts. These tools offer a dynamic and engaging experience while optimizing prompt outputs.

• PromptPerfect: A state-of-the-art prompt optimizer tool that enhances prompts for various types of language models, including ChatGPT, Midjourney, DALL-E, and StableDiffusion.

• Midjourney Prompt Helper: A prompt generator for converting text to image, crafted exclusively for Midjourney and DALL-E. It focuses on ease of use and accessibility.

Courses

• Prompt Engineering Guide: A GitHub repository serving as a treasure trove of resources for prompt engineering, including learning guides, scientific papers, blog links, tutorials, and datasets. Valuable for both developers and practitioners.

• Learn Prompting: A platform offering extensive prompt engineering documentation, courses, and a Discord server with over 1,000 members for exchanging ideas, finding collaboration partners, and staying updated.

• Prompt Engineering Institute: Offers both technical and non-technical courses.

• AI Academy: Offers a Master in Prompt Engineering, aimed at learning advanced prompting skills and how to use prompt management tools to validate prompts' performance and reliability. Also allows to build your own generative AI project (without any code) and add it to your portfolio to showcase your new skills.

Prompt engineering has the potential to reshape the future of work by improving the way humans interact with AI systems. Much like we take courses to improve communication at work to be more efficient in our meetings and with our colleagues, we should start to do the same for how we instruct machines (or ‘Co-bots’ as I called them in my previous article ‘💫 Co-bots: how to think of the synergy between Humans and Machines).

📖 What I’ve been Reading:

• Metaverse Monday is a weekly newsletter that teaches you about the AR/VR industry by interviewing the founders, investors and developers building it. Join readers from Google, a16z and Nvidia.

• AIModels is a weekly newsletter aimed at deep-diving on the plethora of new AI models, with new features explainers and so on. For the engineers, but also for the curious.

• HR Brain Pickings a great career and HR newsletter covering the latest in layoffs, hiring, HR, and misc. career news.