Generative AI is having a moment. ChatGPT and art generators like DALL-E 2, Stable Diffusion, and Midjourney have proven their potential, and now millions are racking their brains over how to make their output resemble the vision in their heads.

This is the goal of prompt engineering: the skill of making an input to deliver a desired result from generative AI.

Image created with Midjourney. Prompt: oil painting of a child with his grandparent enjoying a moment together, looking at each other. The child’s face is full of wonder, and the grandparent’s face is marked by years of life, nostalgia, happy and sad memories, and the wisdom of their years. Detailed faces. – – ar 3:2 – – no glasses.

Despite being trained on more data and computing resources than ever before, generative AI models have limitations. For example, they are not trained to produce content aligned with goals such as truth, insight, reliability, and originality.

They also lack common sense and a fundamental understanding of the world, which means they can generate flawed (and even nonsensical) content.

As such, rapid engineering is essential to unlocking the capabilities of generative AI. And luckily it’s not a technical skill. It’s mostly about trial and error, and keeping a few things in mind.

ChatGPT

Let’s use first ChatGPT to illustrate how prompt engineering can be used for text output. If used effectively, ChatGPT can generate essays, computer code, business plans, cover letters, poetry, jokesand more.

Since it’s a chatbot, you might be tempted to start a conversation with it. But this is not the best approach if you want tailored results. Instead of, adopt the mindset that you program the machine to do a writing job for you.

Create a table of contents similar to what you might 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, target audience and text length
• a list of points you want to cover
• from which perspective do you want the text written, if applicable
• and specific requirements, such as no jargon.

If you want a longer piece, you can generate it in steps. Start with the first few paragraphs and ask ChatGPT to continue in the next prompt. If you are not satisfied with a particular section, you can request that it be rewritten according to new instructions.

But remember: no matter how much you tinker with your prompts, ChatGPT is subject to inaccuracies And invent things. So don’t take anything for granted. In the example below, the output lists a “report” that does not exist. It probably included this because my prompt asked for it only use reliable sources.

I used prompt engineering to get ChatGPT to write this news article, which contains inaccurate information.

Art generators

Midjourney is one of the most popular art generation tools, and one of the easiest for beginners. So let’s use it for our next example.

Unlike text generation, verbose prompts aren’t necessarily better for image generation. The following example shows how a basic prompt combined with a style keyword is enough to create a variety of interesting images. Your style keyword can refer to a genre, art movement, technique, artist, or specific work.

The following images are based on the prompt leopard on tree followed by keywords in different styles. These were (clockwise from top left) synthwave, hyperrealistic, expressionistic And in the style of Zena Holloway. Holloway is a British photographer known for capturing her subjects in ethereal and somewhat surreal scenes, usually underwater.

Images generated by Midjourney.

You can also add keywords related to:

• image qualities, such as “beautiful” or “high definition”
• objects you want in the picture
• and lighting and colors.

Midjourney even lets you use some specific commands for different functions including ––ar or –aspect around the aspect ratio, –No to omit certain objects, and –C to produce more “unusual” results. This command accepts values ​​between 0-100 after it, with the default value being 0 and 100 leading to the most unusual result.

You can also use –S or –stylize to generate more artistic images (at the cost of following the prompt less closely).

The following example applies some of these ideas to create a fantasy image with a dreamy and futuristic look. The prompt used here was dreamy futuristic cityscape, beautiful clouds, interesting colors, cinematic lighting, 8k, 4k ––ar 7:4 ––c 25 ––no windows.

Image generated by Midjourney.

Midjourney accepts multiple prompts for a single image if you use a colon. This can lead to results like the image below, where I’ve given separate prompts for the owl and plants. The full prompt was oil painting of an ethereal owl :: flowers, colors :: abstract :: wisdom ––ar 7:4.

Image generated by Midjourney.

A more advanced type of prompting is to include an image as part of the prompt. Midjourney then takes the style of that image into account when generating a new one.

A good way to get inspiration and ideas is through the Half way down the gallery And style libraries.

Despite amazing results, generative AI is subject to inconsistencies, such as the floating branch in this image. Prompt: woman watching the sunset, magical realism, very beautiful, nature, colorful, very detailed – – ar 7:4

A career of the future?

As generative AI models enter everyday life, more enabling skills are likely to emerge wantedespecially of employers looking for results using AI generators.

Some commentators asking if becoming a “prompt engineer” could be a way for professionals such as designers, software engineers and content writers to save their jobs from automation by integrating generative AI into their work. Others have suggested prompt engineering will be a career itself.

It is hard to to predict what role prompt engineering will play as AI models progress.

But it’s almost a given that more sophisticated generators will be able to handle more complex requests, inviting users to expand their creativity. They are also likely to better understand our preferences, reducing the need for tinkering.

This article has been republished from The conversation under a Creative Commons license. Read the original article.