Introduction
Recently, I have been fascinated to witness the proliferation and experimentation with image-from-text services driven by AI technology. After so many years of feeling secure that our creative talents were one of the few realms of human ability that would remain beyond the reach of machine learning, the output from platforms like Midjourney and Dall-E can feel quite confronting.
For the uninitiated, these platforms generate unique images from a series of written prompts that describe what the user is trying to accomplish. By simply typing what you want to see, the interface generates a series of images as options. You then select what you want to progress and can refine the output until you’re satisfied with the image or set of images.
What does this mean?
The ability to generate striking and emotive imagery shows a fascinating replication of composition, light balance and colour theory produced without human intervention. Themes that can take years to master for an artist can now be bypassed with intelligent prompt inputs, where creative input shifts from production to curation.
Suddenly, everyone can have access to the ability to create captivating imagery with the input of a few words. So, how does this disruptive technology impact designers and creative professionals, and what role can it play in the design process?
I trained as an architect and when I graduated, was fortunate enough to do what I loved, designing a broad range of buildings and spaces with a large degree of creative freedom. It’s worth noting that this is not common in our industry. Architecture is a profoundly misunderstood profession and demands much more than just design. I was lucky because I had acquired a set of skills that allows me to explore my ideas and communicate them persuasively and then landed in an environment where this was encouraged.
My focus for the last several years has been working with other architects and designers to explore, understand and communicate the potential of their ideas using the skills I had developed to express my own. So I approach this new technology from the perspective of a designer, communicator and artist.
I believe that the methods we use to communicate and develop our ideas play an important role in the quality of the outcomes. A rough sketch is still sometimes the best tool to communicate an idea, even when we have access to photorealistic renderings or virtual reality. The appropriate method of exploration or communication is linked to resolution and imagination. The lower the resolution, the higher demand is on imagination to understand the intention. Finding the balance between these is a skill, but you also need to have the option to choose.
Jochen Weißenberger - School of Jelly
Benefits
To varying degrees, every design process has creative and pragmatic stages. This is more true with projects that require translation to functional reality than purely digital or artistic projects, but they all require ideas at some point. It is in the most creative stages of the design process that this technology has the most potential benefit.
Ideation is normally the very start and the most creative stage; it is about throwing ideas out there and discarding most of them. Many designers refer to their process as iterative, which draws an interesting parallel to how users interact with prompt-based image generation. Design is a process of continual refinement; any process that helps generate more thought-provoking ideas in these stages certainly helps.
These new tools have the potential to be an excellent cure for the blank page. Any designer will be familiar with the equivalent of writer's block. Architects are trained to switch mediums or perspectives to help escape these ruts. Whether it is drawing in section rather than plan or creating physical models instead of drawing by hand, understanding what our options are and acquiring more makes us better designers. The more modes of exploration we give ourselves, the better we can understand the problem we're trying to solve.
This is where I see these platforms being powerful: as alternate tools for producing novel expressions based on fuzzy notions or moods. Inspiration can come from much stranger places than artificially generated imagery. Inspiration can strike or be desperately sought at every stage of the design process. Sometimes just seeing something outside of the tunnel vision that can develop is all that is needed.
There are, of course, many applications outside the design profession. The African Institute for Artificial Intelligence (Ai_Ai) generate images to highlight and explore the inherent bias in machine learning systems so that these can be discussed and bypassed. With relatively minimal input, they can generate provocative imagery that prompts discussion and reaction in a community that aims to ensure that any creative representation doesn’t simply perpetuate stereotypes. This highlights one of the main risks of machine learning and also calls into question the accuracy of Artificial Intelligence as the latest buzzword. There are many examples of these biases in other systems and text-to-image generation is no exception.
Challenges
However, opportunity rarely comes without risk, and these services could easily be mistaken for something they’re not and represent a potential pitfall for young designers. Lured by the promise of beguiling imagery, I imagine architecture students spending hours inputting prompts and waiting for the images they hope will come but may not. If they do, how do they explain the thinking that led them there? Or worse, how will they transform this into reality?
This touches on another issue: using these images at the wrong project stage. Expectations are placed on architectural visualisations to capture the feeling of being in a place while accurately depicting the reality of what will be built. There has been serious debate around the manipulation of truth through visualisations to misconstrue how a project will appear upon completion. What happens when the process of communication is completely detached from the design? Reverse engineering art, although not impossible, is a path normally reserved for the pioneers.
Conclusion
AI has the potential to help jog us out of our normal patterns of thinking by asking something rather than somebody else what they think. With the responses being so beautiful, it’s easy to believe this input will inspire designers to greater heights of imagination.
The challenge will always be in translating imagination into reality, but that is what great design is.