Status of generative AI in architecture and planning

Artificial intelligence and its potential applications are currently THE hot topic. There is daily media coverage on all channels, and AI is probably one of the most frequent topics of conversation at office coffee machines in Germany. Companies specializing in the use and operation of AI are traded on the stock exchange at astronomical prices. Fake images and deepfake videos created with the help of AI are causing crises and straining political relations at the highest diplomatic levels.

Few, however, realize that they are already users of AI-supported applications in their everyday lives, or even proactively work with corresponding solutions.

Powerful AI applications are now becoming more and more commonplace. Generative AI tools in particular are currently attracting a great deal of attention, as they can produce completely new and creative content. We are both fascinated and shocked by how good the results of these tools have become. Especially in the field of creative professions, humans have always felt vastly superior to machines, but even this domain is now being increasingly and successfully conquered by AI.

Open AI’s ChatGPT software is currently dominating the headlines rather unchallenged, as it is publicly available free of charge and can be used to generate any text of high quality and quite some originality. At the moment, however, AI applications often still have teething problems, and developers are struggling to scale their tools to millions of new users.

For architects and planners, such text-based tools are currently more of a marginal topic for their work, as they mainly deal with visual and technical issues. However, there are already some remarkable AI applications on the market that support visual creativity.

Microsoft recently integrated Microsoft Copilot, an AI technology based on Open AI solutions, into its Office suite, which can also generate images from text input, among other possibilities.

One of the leading companies in the field of creative software applications, Adobe, is currently investing heavily in its own AI technology with Adobe Firefly for seamless integration into its own creative suite.

In addition, the software Midjourney, which is accessible via a chat platform called Discord, is a very interesting AI image engine that is based on its own proprietary, high-performance AI technology.

 

How do these applications work?

 

Users interact with tools such as Midjourney to generate the desired image motifs based on the input of text commands that are as precise as possible, so-called “prompts”. Here, the user can describe in text what content an image should contain and what type of display is desired. The amazing thing here is that the AI applications are perfectly capable of processing several desired image elements into a coherent, realistic-looking new whole. For example, if you want a picture of a modern house with a wooden façade on the shore of a lake with a forest in the background and finally the whole thing in a wintry atmosphere, the AI engine generates realistic image motifs in the desired combination of these six individual key pieces of information without any problems, even though the individual elements have to be interlinked and cannot simply be lined up next to each other in an additive manner. Of course, the software sometimes generates image motifs that do not directly correspond to the desired result. In most cases, however, a satisfactory result can be achieved by varying the text input. If there are too many individual desired elements, or if the desired image content is contradictory, a coherent and usable image will not be created. However, an image fail is often also due to the user’s imprecise idea of the desired target motif. In view of powerful and efficient tools, our own imagination increasingly sets the limits. In some cases, however, the AI engine unexpectedly generates nonsensical pictures or incorrect details even for supposedly easier tasks, as the underlying technology is extremely complex and not yet fully developed.

Sophisticated creative professions will not be replaced by AI solutions in the foreseeable future, but it is fair to predict that the creation of visual works with the support of artificial intelligence will become a permanent fixture in the future, and that users who proactively embrace these opportunities in good time will gain a strong productivity advantage. Especially in the creative early phase of projects, AI tools can quickly and easily visualize new ideas as well as expand one’s own horizons. AI-supported solutions could also soon be used as a matter of course when compiling mood boards, as the results can be quite unexpected in a positive sense and are useful for one’s own discovery process.

In addition, the new AI possibilities will not go unnoticed by clients, and planners may find themselves confronted with the fact that their clients are now approaching them with even more detailed and complex visualizations of their wishes. As the AI-based visual applications will foreseeably not yet have any testing routines with regard to the technical feasibility of certain spatial representations, this may also lead to new problems in the future in the joint definition of the desired goals between client and planner. For this reason alone, creative professionals should be aware of the main application horizon as well as the current limitations of generative AI tools.

However, companies in commercial image databases such as Shutterstock, Getty Images and others will find it very difficult in the near future to charge comparatively high prices for exclusive images or their usage rights, as users are likely to increasingly fall back on alternative solutions based on AI models that are optimally tailored to their individual needs and generated quickly. For the photographers behind the databases, this development could also reduce previously established sources of income.

But not only new images can be created with the help of AI tools. Applications based on artificial intelligence can also be used to modify existing image motifs. Here too, creative service providers could potentially lose a few jobs in the future. It is no coincidence that Adobe is currently working flat out on its own natively integrated AI solution in order to develop new profitable business models before previously established sources of income are lost to the new competition. One man’s gain is always another man’s loss.

Photographers, artists and producers of creative visual works in general rightly criticize, for example, that AI applications are being trained with the help of their own previous creations without their explicit permission – let alone the payment of license fees – in order to potentially jeopardize part of their business basis in the future.

As far as image rights are concerned, there is currently light and shade and still some gray areas. As a regular user of commercial AI software, you generally receive the rights to use the motifs and can use the generated works commercially without restriction. However, the visual works created with the support of AI are currently not yet subject to copyright protection and can therefore theoretically be copied by third parties at any time and used for their own purposes. Corresponding existing rights regulations will certainly be adapted to the new applications as AI tools become more widespread.

It will be interesting to see how quickly and in which direction generative AI applications will develop in the future. However, one thing is already clear: they are extremely powerful. With the still young start and the manageable market of the first commercial tools, it is now a good idea to get to grips with the new possibilities at an early stage, get to know the potential and limits of the new technologies and make the best use of them.

 

“Chuck Norris doesn’t use AI because even his shadow is smarter than any machine!” (Source: ChatGPT)

Written by: Andre Flinterhoff