By Andi Walter
A closer look at how AI tools’ understanding of light doesn’t always render accurately.
Photo: AI generated from Midjourney
Ten years ago, the concept of artificial intelligence (AI) was something that you might see in a dystopian sci-fi movie or hear tech executives pontificate on as they sold you software. From the conceptualization of AI in the 1950s, AI always felt like a distant future possibility that would not affect our lives any more than teleportation or warp-speed travel. However, in the last few years, AI has quickly emerged as one of the trendiest topics in modern society—not just for big tech companies but also the design industry.
Perhaps you’ve heard of ChatGPT or DALL·E, noticed an “AI assistant” taking notes in a virtual meeting or heard that your building information modeling manager is using it to help streamline the drafting process. The newest iPhone has multiple built-in AI features, and it seems that every company has AI chatbots answering customer questions. Some large firms are even hiring AI-specialist teams in-house. The 2023 Writers Guild of America’s strike highlighted the urgent need to protect artists’, storytellers’, content creators’ and designers’ jobs from being replaced by AI-generated content. The innovations in the world of AI have been swift and are changing every day, so it is crucial for interior designers, architects, engineers and lighting designers to define their goals and boundaries for using AI in their practice.
Before we jump into the benefits, risks and complex considerations for using AI in lighting design, it is important to remember that AI isn’t actually “smart.” Most of the consumer-facing tools you have likely used or heard about are considered “generative AI,” meaning they are generating content based on trained associations found in huge data sets of text and images, including much of what is publicly available on the Internet. AI tools aren’t logical and can’t think creatively or imagine new realities. They only aggregate, reference and organize information that already exists. This can be a valuable tool, but you can likely already see how lighting design doesn’t fit neatly into their framework.
The unique, creative and collaborative world of hospitality projects are a perfect place to begin to understand AI best practices. We live in a hyperconnected world, where every restaurant, hotel and resort is seeking opportunities to provide memorable, novel and shareable experiences. Hospitality clients are looking to draw in visitors with something they have never seen before and hoping it goes viral. So, if we turn to AI tools, how can they help meet these clients’ unique needs?
Beginning with the conceptual development phase of a project, using AI image generation is a great exercise to force you to precisely describe what you’re imagining. The results of a Pinterest search for “modern tropical restaurant” might return plenty of similar ideas that partially capture the concept in your mind. With an AI tool like Midjourney, you can become fluent in writing prompts that generate robust, detailed renderings. By prompting it with “A modern restaurant with group seating, large tropical plants, natural light and colorful décor,” you will generate an image that is influenced by—and compiled from—a whole universe of similar images. It is possible to visualize your ideas quickly and iteratively and create many low-effort options to consider. You can also use this process to begin to describe a custom fixture you would like to have fabricated to fit a particular space or merge several concept images together to communicate a specific style. All these processes can be significant time-savers and help you create a library of your own unique concept imagery. If your concept image isn’t quite right, you can now use Photoshop’s quick generative AI tools to enhance and better convey your idea. There is even a plug-in for Revit that can overlay AI-generated lighting on top of 3-D views without ever leaving the program.
The learning curve for using AI image-generation tools can yield some unexpected results, especially for lighting designers who use technical terms and industry-specific jargon. Take, for example, a request to show a grazing effect on an exterior wall. Our practice’s initial tests resulted in a wide selection of beautiful images of cows grazing next to brick walls. Over time, you’ll learn the logic of the prompts and how to be specific to achieve the desired results. But even with a precise description, many AI-generated images will return ghostly pendants hovering in the air or light with no apparent source. The longer one analyzes these images, the more signs become apparent that they are aggregating rather than designing.
Still, there are some good arguments to be made for the use of image-generation tools. They level the playing field to allow more accessibility for designers with a variety of skill sets to create high-quality renderings. They can help bridge a communication gap between the design team and owners, and the images may spark interesting conversations as part of a charrette process. Their cost is much lower than licensing for professional-grade rendering software, and their speed is unmatched. As designers, the prospect of AI replacing our expertise can be anxiety-inducing, especially when we see photorealistic designs proliferating so quickly into the industry. However, two big things stand on the side of the designer as we face the opportunities and threats of AI technology: basic human empathy and basic physics.
For any designer who has built a new relationship with a developer, owner or architect, it is evident that AI will not replace the emotional intelligence, creative vision and empathy that humans cultivate throughout our projects. AI tools might help you write an email, but they can’t expertly navigate sensitive discussions with clients or colleagues to find the best solution for everyone. AI can’t consider the environmental ramifications of its specifications, the acoustics and materiality of fixtures or the cultural nuances or social contexts within which we design. As experienced professionals, we build interpersonal connections and trust with clients and mentor the next generation of lighting designers. While AI can be a handy addition to a designer’s creative process, it’s still up to us to contribute holistic design thinking, make ethical decisions and use our artistry for the overall betterment of the end user. AI is a tool, but lighting design is an art.
Precise language is required when prompting AI tools with industry terminology. This image resulted from the prompt of “exterior brick wall grazing effect.”
Photo: AI Generated from Midjourney
The physics question is more technical but no less important. Designers know that light has profound impacts on our psychology, physiology and experience of a space. Even professional rendering artists can struggle to communicate accurate light, shadows and reflections that represent a true-to-life feeling in their graphics. Experienced lighting designers know how to communicate their intentions and reassure their clients that the desired mood and function will be possible to achieve. We’ve all seen AI images with uncanny faces, unusual numbers of fingers or nonsense text on signs. AI renderings used to illustrate lighting are just as likely to show something that is impossible to create in real space, risking a client falling in love with an AI image that they’ll never be able to achieve. While it can sometimes drive innovative solutions, most projects don’t have the resources to solve these problems of physics.
After taking these limitations into account, you may decide to start trying out AI tools in your firm. Here’s some best-practice tips learned through an exploratory year of adopting AI into our practice.
Intentional policies and legal implications: Be sure that your firm has an AI policy in place before rolling out any tools beyond a test group. The firm could maintain a list of trusted tools and confirm their SOC 2 compliance, a third-party source for protecting sensitive data. Remember that any image or text you upload into an AI tool for it to alter can become a part of that vast data set on which AI is trained. If you don’t own the rights to a rendering, you shouldn’t upload it to play with AI tools. Similarly, any proprietary drawings, language or even personnel headshots can be tricky areas. It is important to respect the work of other artists and share their work appropriately.
Open communication: When using AI-generated images in a presentation, make sure your team has a consistent method of crediting and sourcing to avoid confusion. Help your team to identify AI-generated images and work transparently to understand the limits and benefits. This can be a fun opportunity for friendly competition to have your current process race against the AI generation and compare the results. Also, remember that if AI is trained on data that is biased toward or against a particular group, it will reinforce those biases, so make sure you’re thinking critically about the solutions it offers to maintain an inclusive design process.
Consider the environmental and human impacts: For many of us, our firms openly value sustainability and social justice, signing on to commitments like the Lighting Advocacy Letter, AIA 2030, or MEP 2040. One piece of the AI puzzle that is coming to the forefront of the conversation is the environmental impact of the amount of computing power needed to drive these technologies. Opting to use AI tools means we’re contributing to the huge energy and water needs of data centers’ computing and cooling. According to The Los Angeles Times, “A query on the chatbot that uses artificial intelligence is estimated to require at least 10 times more electricity than a standard search on Google.” In addition, many AI models are trained on data that has been manually sorted and labeled by exploited workers around the world. These are big ethical conversations to have with your leadership as you explore the use of AI and consider your overall environmental impact through other sustainable design practices and community involvement.
The bar area of Infinite Labs in Louisville, CO.
Photo: David Lauer Photography.
An attempt to recreate the same lighting effect in Midjourney with the prompt, “open daylight cool-toned bar area with arches in rear, contemporary mid-century style with modern decorative pendants.”
The decision about how to use AI as a designer requires us to call on our most human experiences. Passionate designers already use a wide variety of tools to develop and communicate their ideas. Much like during the transition from hand drafting to CAD, our industries are continually evolving. We have a responsibility to determine the benefits and risks of these tools and decide for ourselves if there is something intangible at the heart of the process that is gained or lost by using them. And we have an opportunity to share the art and humanity of design with a new generation who are growing up surrounded by AI and encourage them to understand the kind of expertise that can’t be replaced by technological advancements.
THE AUTHOR | Andi Walter, LC, Member IES, IALD, LEED Green associate, serves as a senior lighting designer at AE Design and has a passion for innovative, sustainable and collaborative design.