Advancing artificial intelligence
The AI is already shaking the foundations of visual art and raising questions about artistic democracy, authenticity, ownership and censorship. In the world of architecture, the challenges are slightly different. As yet (until this article was written), there is no direct implementation method for current practice other than using AI tools (such as Midjourney, DALL·E 2, and Stable Diffusion) as a rapid conceptualization/outlining medium that expands architects’ and designers’ fantasy.
Hassan Ragab is an interdisciplinary designer with an architectural and computational design background who is very interested in visual arts. He reviewed his experience using AI text-to-image generators for Parametric Architecture, which started about two months ago, focusing on Midjourney as the primary tool. He currently uses it to create digital art.
What is Midjourney?
Midjourney is an AI text-to-image diffusion model hosted on a Discord server. It works by entering text descriptions or prompts. Primary settings like these prompts (aspect ratio, how creative the AI can get by ignoring parts of your text, etc.) can be adjusted. The AI bot will then take less than a minute (depending on the mode you’re on and server status) to produce four variations. You can then expand on any of these four results by scaling up or creating more variations (of course there are more technicalities to spur craft that everyone is still exploring).
According to Hassan Ragab, Midjourney is a fascinating tool for exploration for the following reasons:
- Expanding the limitations of architectural inspirations by fusing architecture with artistic surrealism and abstract movements in a free domain away from the physical constraints.
- Heritage as a source of inspiration. It is important to diversify our understanding of architecture in the era of futurism – both in the natural and virtual worlds. He believes that an invaluable visual awareness can be achieved by mixing traditional schools of architecture with modern art movements and modern design vocabulary such as computational design aesthetics.
- Using this tool as more than just a sketching and drawing tool: Other than the democracy that these tools imply, there are deep processes within themselves. These processes are perceived as enhanced extensions of our understanding of creativity, where all art is connected through an endless process of a “parametric copy-paste”. The prompts you use, your variant selection decisions, and your ratings are all cornerstones of how AI models are trained. As a consequence, these decisions are not only fed back to you; but to everyone who uses the tool. So we build all these tools together as we take a massive part in creating our collective imagination.
Having a solid opinion about this tool is such a complex endeavor; to be in the middle of exploring a tool that is itself in its infancy and developing at such an unprecedented pace is no easy task.
Here’s what Hassan Ragab likes and dislikes about it so far:
- The artistic value: Unlike Stable Diffusion and DALL·E 2, Midjourney models “except for the latest V4 test” are tuned to favor “beautiful and artistic” results over photorealism. He finds it a meaningful way to work with artificial intelligence; is to use the true power of Ai in your creative process as an extension of your imagination instead of trying to come up with realistic rendering that could be done using other software. In order not to undermine the ability of these tools to generate realistic renderings in a matter of minutes as a means of architectural presentation, he sees the latter as a low-hanging fruit rather than the essence of working with such tools.
- Speed, mobility and accessibility: You sketch your ideas on the fly, faster than any human before. All you need is a Discord app at your disposal and you’re good to go—another key element in how this tool amplifies human creativity.
- Smart conversation: He has always imagined art and design as a conversation between the creative and the medium – it is not one-sided, but a beautiful, slow dance. This is the experience that AI evokes as it enriches artistic concepts in ways never imagined before.
- Community: MJ developers organize open office hours every week. In these meetings, they explain their processes, business model and future goals, while opening the discussion floor for literally all users to debate, discuss and recommend. In addition, there have been good networks between designers, architects, artists and even non-professionals that are being developed all over the world. This kind of community involves everyone in holistic and inclusive conversations and collaborations on social media (especially on Instagram). The effort to discover this new world makes you believe – in the midst of everything that is happening in the world – that the different people from different cultures can stick together and form a complex intangible entity where the sum is more significant than its parts.
Although the current Midjourney models lack personal styling tools. Work is ongoing to create a holistic model that combines the functions of all the models. As seen in their latest V4 model, which focuses more on creating photorealistic results “probably through macrosampling”. However, one problem with the new model is that it removes a lot of the “artistic vibes” from the image. The latest remaster tool also allows restoring V3 generated images using V4 new capabilities. Hassan Ragab mentioned that he doesn’t like the V4 model. He believes that the models will be developed between creativity and photorealism. It’s also worth noting that other AI tools offer smarter post-processing features. An example is the new Outpainting tool in DALL·E 2. He had the chance to try Outpainting in Midjourney outputs. He mentioned that he was really impressed with the result.
- According to Hassan Ragab, who works with any AI text-to-image, it is a precedent for any creator to implement so many ideas in such a short time. These influences make users more picky about results than ever before, resulting in frustration after the euphoria the user gets as a beginner. We are still learning to adapt to this new medium and it may take a while to fully understand the current reality.
- Since the AI generators are the democratic tool (available to anyone willing to use it), there are just too many ideas out there that are just too good. Too many people develop good concepts in such a short time, which sets the bar higher for the minute. While this can create the environment for a healthy competitive field, it also puts some pressure on the AI creators. They feel that they must always be aware of each day’s developments and execute their ideas as quickly as they can before someone else can beat them to it.
- As he mentioned, the Midjourney models learn from us and feed back the results. As time goes on, this began to manifest itself in the emergence of some “universal/apocalyptic style”. This is also due to the lack of customization capabilities in Midjourney and the limitations of customizing the model to your style. He finds this a decisive flag. Not taking this seriously can jeopardize the diversity and individualism that art is all about.
- Lack of diversity in the dataset for non-Western architectural landmarks may lead to the lack of demographic diversity of the users. Both can drive this technology away from inclusivity and diversity. That’s what he doesn’t like about working with Midjourney. As an Egyptian, he always finds excellent architectural inspiration from his homeland. Egypt has one of the richest architectural histories in the world, but he struggles to use Egyptian landmarks and cities in a meaningful way. The bad data sets of certain places; architectural and artistic styles are significant constraints for many creators in Egypt and worldwide. This limitation may drive some users away from these techniques to explore other methods with extensive representation in the model dataset, such as the work of Zaha Hadid.
To conclude, regardless of the ongoing debates and the apocalyptic flags, he believes that AI tools are here to stay. It is already changing the creative world as we know it and knocking loudly on the door of architectural practice. An excellent race to develop these tools is taking place at an incredible pace. Works of art attract almost everyone, change what we know about creativity and make many acquired skills obsolete. He is hopeful about how these tools will change us for the better – these tools are seen as double-edged weapons, requiring us to be considerate.
According to Hassan Ragab, we should be critical of our ways when dealing with said tools. We should always keep an open mind to the new while being aware and prepared for the coming shift or we will perish as creators.
About Hassan Ragab
Hassan Ragab is an Egyptian multidisciplinary designer with an architectural background living in Southern California. His experiences vary between architectural construction, exhibition design, product, furniture design, textile design and digital art. His work is better connected to computation, experimentation, digital fabrication and exploration of visual relationships between architecture, heritage and art through artificial intelligence.
Hassan Ragab gives a lecture at Computational Design: NEXT 11 on 8-9th October 2022. Join the program to learn more about Hassan Ragab’s works and much more!
Computational Design: NEXT 11 brings together pioneers and professionals from all industries to meet and discuss new emerging technologies in art, architecture, design and manufacturing.