How to tackle with AI in AEC?

by | sij 12, 2024

Considering the construction industry is sluggish, particularly in adopting new technologies, it’s at the very end of the line. As we’ve seen with BIM (Building Information Modeling), it takes us a long time to adopt more advanced solutions. I believe the same will be true for AI implementation. On the other hand, this could be an advantage for the agile ones to exploit this situation and, as with BIM, leap ahead to grab a larger piece of the pie.

However, the use of artificial intelligence, while bringing numerous advantages, also raises some questions. Maybe it’s good to start thinking about this in time. Because when these situations arise, there likely won’t be time. Here are some of them…

Who/What is behind this design?

In the beginning, AI tools will be those that use AI technology to a lesser extent for some parts of the process, so this question won’t be immediately relevant. But over time, AI may handle a good part of the creative process.
The question that might arise – do we need to know if AI is the creator or co-creator of a design? Or does it not matter, and the only important thing is that the project is satisfactory? That is, is only the result important or the process by which the result was achieved?

Why it might be important whether AI participated in the design solution we will see through the following questions.

Who is the author?

If an AI algorithm “creates” a design, who is actually the author of that design? There are many actors involved in this process. The company owning the algorithm, the content on which the algorithm was trained, the algorithm itself (whatever that may be), or the designer who initiated it. Who among them can claim authorship? What will be the original in each project that will determine who is the author: The idea? Describing that idea (today in the form of a prompt, and tomorrow who knows how)? Shaping that idea into a result? The way (algorithm) in which a simple idea is shaped into a result? The content used to achieve that result. Is there only one author? Are we all co-authors? Does it matter who is the author? …

Inspiration, reference, or plagiarism?

Architecture certainly does not reinvent the wheel on every new project. Similar elements are used in all projects, and generally, buildings differ only in combinatorics and expression. Likewise, references that inspired us or on which we “leaned” are often cited in project descriptions to give our projects more “credibility.” I’m not always sure what the exact difference is between referencing and copying someone else’s solutions. Sometimes it just seems to me that it’s only in the semantics of interpretation. 😊

How will we view this in the context of the originality of architectural parts that AI will create in the future? It will also use some architectural works as a “model” (it will be trained on them). Will we understand this as references, or will we treat the obtained design as plagiarism? Is an AI-generated solution original if we know it consists of parts of other solutions, or is it a copy? In the music industry, there is a defined measure that should not be exceeded in similarity to an existing song for it not to be plagiarism. Do we need something like that in our industry?

Quality of AI solutions?

Following the previous question. If we adopt that it’s okay for AI to use some existing projects for “inspiration,” the question is how high quality were those examples? What database of projects will AI use as a reference? Are these some relevant global examples or just anything that could be collected from the internet? Will we know what was used if we used such tools? So far, it’s all based on photographs or illustrations, but I believe that in the future there will be AI tools that will “penetrate” much deeper and give us more complete solutions.

In the world of photography, this is already a real situation today. A huge number of images used for various commercial purposes are generated with the help of AI. In a way, we do not know the source on which AI was trained. However, some tools use a precisely defined and controlled database of images from which AI generates new images. Thus, at least some problems are solved – quality and compensation to the original authors. Certainly, some AI solutions will be of higher quality than others and will become increasingly higher quality over time. But who is the one who will decide if the solution is of sufficient quality? Will there be reviewers of AI-generated design solutions?

AI “confidentiality”?

If we generate a solution using an AI tool, to what extent can we be sure that only we have knowledge of that design? If we are working on a project that requires secrecy – do we have guarantees that this solution has not been stored somewhere else or that someone cannot reconstruct it by following a similar process of obtaining that solution – simply with a similar prompt? Can we guarantee to the client that the project is only in our and their archive?

AI as an equal subject in the process?

Is AI just a tool in the hands of the designer, or is it a subject that “itself” creates design solutions? When we approach the latter, will such a solution be equal to a solution obtained by the “classical” method? In our industry, this is still a bit hard to imagine, so I’ll give an example that is already real today. If AI submits a song for Eurovision – can it compete?

Time is now

I don’t think we need to have answers to these and some new questions that will arise today. Nor do I think this is a topic we can solve locally, at least for the most part. But I think it’s time to start asking questions. And to try to cover as much as possible so that everything is as clear as possible to us when this becomes a reality. Because all this with AI can still go in various directions.

Without universally recognized and prescribed frameworks, we may start to get cheap design solutions, irresistibly looking, questionable quality, and maybe even plagiarized, which will be unfair competition to higher quality original solutions. But since most “ignorant” investors will evaluate which to implement, the question is which will have the advantage?

On the other hand, the use of AI can lead us to phenomenal, finely optimized original solutions that are adapted to the required purpose in a way and form that we would hardly reach in that time frame and budget by conventional means.

And how it will really look in the future, we will all have to somehow agree. And I hope we do it in time.

Author: Vedran Orešić

Kao BIM konzultant i Autodesk Certified Instructor pomažem kolegama da implementiraju napredna BIM rješenja kako bi svoj svakodnevni rad učinili efikasnijim.