Generative AI is currently experiencing a surge in popularity. ChatGPT and art generators like DALL-E 2, Stable Diffusion, and Midjourney have showcased their potential, prompting millions of people to strive towards realizing their desired output through these platforms. [BOLD]
Prompt engineering aims to use the art of formulating inputs to obtain the desired output from generative AI.
Despite being trained on vast amounts of data and computational resources, generative AI models still have limitations. One of their weaknesses is that they are not designed to produce content that aligns with truth, insight, reliability, and originality objectives. Moreover, they lack common sense and a basic understanding of the world, which can lead to the generation of flawed or even nonsensical content.
Hence, prompt engineering is crucial for unleashing the full potential of generative AI. Fortunately, it does not require extensive technical skills. It mainly involves trial and error and keeping a few key considerations in mind.
ChatGPT
Let's consider ChatGPT as an example of how prompt engineering can be used to generate text outputs. With effective prompt engineering, ChatGPT can produce various text forms like essays, computer code, business plans, cover letters, poetry, jokes, and more. However, since it's a chatbot, you may be tempted to engage with it conversationally.
But for tailored results, this isn't the best approach.
Instead, consider that you're programming the machine to complete a writing task for you.
Prepare a content brief similar to what you would give a hired professional writer.
The key is to provide as much context as possible and use specific and detailed language. You can include information about your desired focus, format, style, intended audience, and text length, a list of points you want to be addressed, what perspective you want the text written from (if applicable), and specific requirements such as avoiding jargon.
If you want a longer piece, you can generate it in steps. Begin with the first few paragraphs and ask ChatGPT to continue in the next prompt. If you're unsatisfied with a specific portion, you can ask for it to be rewritten according to new instructions. However, remember that no matter how much you tweak your prompts, ChatGPT can still produce inaccuracies and make things up. Therefore, don't take everything it generates at face value.
For example, the output below mentions a "report" that doesn't exist. This may have been included because the prompt asked ChatGPT to use only reliable sources.
Generating Art with AI
Let's talk about art generators, with Midjourney being one of the most popular tools for generating art and one of the easiest for beginners. Unlike text generation, complex prompts may not be necessary for generating images.
The following example demonstrates how a simple prompt combined with a style keyword can create various intriguing images.
Your style keyword may refer to a genre, art movement, technique, artist, or specific work. For instance, the following images were generated based on the prompt "leopard on a tree" followed by different style keywords: synth-wave, hyperrealist, expressionist, and in the style of Zena Holloway.
Zena Holloway is a British photographer recognized for capturing her subjects in surreal and ethereal scenes, most commonly underwater.
You can also incorporate keywords related to various aspects of the image, such as "beautiful" or "high definition," specific objects you want to be depicted, and desired lighting and colors.
With Midjourney, you can use specific commands to control different features, such as "--ar" or "--aspect" to set the aspect ratio, "--no" to omit particular objects, and "--c" to produce more unusual results. The "--c" command accepts values between 0-100, where the default is 0, and 100 generates the most unusual results.
You can also use "--s" or "--stylise" to generate more artistic images at the cost of following the prompt less strictly.
The following example applies some of these techniques to create a fantasy image with a dreamlike and futuristic appearance. The prompt includes the phrases "dreamy futuristic cityscape," "beautiful," "clouds," "interesting colors," "cinematic lighting" and specifies an aspect ratio of 7:4, with 25% unusual results and no windows. Midjourney can accept multiple prompts for one image if you use a double colon (::), leading to results like the image below, which uses separate prompts for an owl and plants. The full prompt for this example was "oil painting of an ethereal owl :: flowers, colors :: abstract :: wisdom ––ar 7:4."
A more advanced way to prompt Midjourney is to include an image as part of the prompt. Midjourney will then consider the style of that image when generating a new one.
Browsing the Midjourney gallery and style libraries is a great way to find inspiration and ideas.
Is Prompt Engineering a Career of the Future?
With the increasing prevalence of generative AI models in our daily lives, skills in prompt engineering are likely to become more valuable, particularly for employers seeking to achieve desired outcomes using AI generators.
Some experts suggest that professionals such as designers, software engineers, and content writers could potentially use prompt engineering to augment their work and potentially protect their jobs from automation. Others see prompt engineering as a career in itself.
It's difficult to anticipate the exact role prompt engineering will play as AI models continue to evolve. However, as generators become more sophisticated, they will likely be able to handle more intricate requests, encouraging users to explore new levels of creativity. Additionally, generators will probably better understand our preferences, potentially reducing the need for manual adjustments.