The introduction of generative AI applications, such as DALL-E and Stability AI, have brought an opportunity to architecture and design to work in conjunction with the computer as designer. The technology creates imagery that is based on text inputs. The number of variations seems endless, and the user can make multiple iterations, each of which does not repeat in any details. The images are each unique but the content is repeated from images reviewed by the application– each is a rehash– there is nothing new in these, beyond what has already been created. As such, the artificial intelligence is not truly creative, however it can be a tool to enhance creativity.
We can treat the images as artifacts to analyze. To do this, we can use the traditional formal analysis used in art history to describe and analyze the work. It is not necessary to think of the images created by the computer as art–formal analysis will still work. Formal analysis looks at two-dimensional and three-dimensional works and reviews them through by developing an understanding of the work’s description, analysis, interpretation, and evaluation. The description is a basic understanding of what the formal qualities of the work are. Analysis looks at how the parts of the image or object work together to create the whole. Interpretation explores the meaning of the work from a point of view. This interpretation can be individual, regional, or shared by everyone. Finally, the evaluation is the summation of the description, analysis, and interpretation. This evaluation can vary among individuals, regions, and the globe, but it is important to do this step to understand how the object and its use compares and contrasts with other viewpoints.
After doing the formal analysis, the designer can have a better understanding of the purpose or advantage of the image in relation to the objective. This allows the designer to look at the objective with fresh eyes, but it also makes the designer review the AI-created image to find how it is different or uncanny. How can the image not be accurate? Why is the imagery wrong? What would make it more accurate? Using the answers to these will help guide the designer to create a fresh or new viewpoint related to the subject. So, the first method of using the Generative AI is to design through what is incorrect with the AI images.
Next, there is an importance to compare and contrast the imagery created with the designer’s are there any advantages to those created by the AI. Of course, any ingenuity is serendipitous for the computer– using a montage of previous images to create a unique image. However, it is still possible for the AI to come upon something surprising and possibly original through juxtaposition. How can we use the ideas provided in the AI images, if there are any? How can we ensure we do not create something similar to that of the AI, if the results are poor? Comparison and contrast will allow us to design more effectively.
The next way we can use Generative AI is to hone the word choice on the generator’s prompt. By editing and developing the selection of words in the input line, we can ensure we are succinct and exact in our description of the desired effect. The tool can be powerful, but it can only produce by following the description and use images that seem to relate to the prompt. So, it is important to have a strategy when entering the text and providing any context, as necessary. The computer tends to follow preconceptions with some terms, so there are similarities between the iterations, and it is helpful to perfect the method of input to limit the bizarre and strange, unless that is the intention and purpose.
Finally, because the Generative AI combines and splices existing imagery, creating a montage, Generative AI can be useful to create partis and explore the thresholds and boundaries that connect the various elements in the prompt. The system is very fast at creating amalgamations of items, and these can get the designer thinking about possibilities, even if what is produced by the application is not realistic or with flaws. Careful examination of the transition between parts can produce novel possibilities and solutions.
There could be other possibilities of using the Generative AI, and these will appear with more use of the application. I can assume the use of plans, three-dimensional imagery, and modifiers will also create some interesting results which could inform design.