Expectations surrounding artificial intelligence have reached unprecedented levels. Back in June 2023, we shared insights on our AI integration into the Scan2BIM process.
At that time, skepticism towards AI prevailed, but fast forward almost a year later, the expectations are higher than ever. During our participation at digitalBAU, we engaged with entrepreneurs actively seeking AI solutions to optimize their workflows. The journey to practical AI solutions presents its own set of challenges, with many of the elevated expectations rooted in the pervasive marketing hype surrounding AI.
However, the question remains: How long do we still have to wait for truly beneficial and effective solutions?
Looking back at June 2023
In the previous year, we delved into the offerings of the "AI market," conducting tests on different plugins and striving to incorporate them into our Scan2BIM workflow.
The conclusion back then? "AI proves to be beneficial, particularly in tackling repetitive tasks, yet the revolutionary advancements in the Scan2BIM domain were not available then, nor are they now," observes Sandu, one of our Revit BIM/CAD experts. However, what further advancements can we anticipate from AI? What remains visionary, what is within reach, and what has already materialized?
Our expectations and demands of AI
When discussing artificial intelligence, it's advisable not to delve into utopian ideals. What was once considered impossible has now become routine. Tesla CEO Elon Musk predicts that by 2026, artificial intelligence could surpass the intellect of even the brightest human minds. This assertion, reported by "Standard" on April 9, 2024, has also been echoed by various other media sources.
"AI serves as a valuable resource for ideas generation. Yet, it's essential to critically evaluate AI-generated information, as only highly skilled experts can grasp and assess the intricacies of a subject. I firmly believe AI can be a substantial asset. However, in the 'Scan-to-BIM' field, groundbreaking tools might require a longer wait," states ArchiCAD CAD/BIM expert, Liudmila.
"We're hopeful that AI can handle simple tasks and workflows on its own, especially in Scan-to-CAD/BIM projects," Michael states. "In less demanding projects, the potential time savings in modeling are significant, and we're optimistic that the benefits will outweigh the costs. Currently, loading a building's point cloud into CAD software and generating the model with a mouse click works best for simpler structures (LOG100-200) and without considering all model requirements*. However, I believe AI will eventually be able to convert more complex buildings into high-quality 3D models with minimal intervention from the modeler."
"Based on our experience, customizing model requirements to suit the future use of the model is essential. This customization process presents the greatest complexity, as the results for the same building can vary significantly depending on the end-user and their specific needs. The CAD (or BIM) model is constructed and detailed differently depending on these requirements," Philip explains. "This poses a challenge for developers and modelers alike: developing and training AI to ensure that the models effectively serve their purpose despite their diverse complexities and individual characteristics."
"We expect maximum efficiency potential from AI tools when they're tailored for specific use cases and excel in them, thereby reducing the need for extensive post-processing and quality assurance."
"Based on current trends, the top Scan-to-BIM AI tools excel in representing basic components, requiring further enrichment with detailed information," Sandu explains from his practical knowledge. "Currently, AI tools serve the purpose of supporting and, ideally, automating repetitive and time-consuming tasks," Michael emphasizes. "For these tools to be cost-effective, they need to contribute significantly to the overall project efficiency," Amy concludes.
What does the market look like, and what does it promise in the near future?
Developing AI tools for Scan-to-BIM is of significant complexity. (Scan2BIM LOD Definition) Modeling from point clouds is a comprehensive and often customized process, requiring both time and financial investment before releasing lucrative and rewarding tools. Fortunately, numerous companies are actively pursuing solutions.Â
For instance, Graphisoft is actively developing various AI tools for ArchiCAD and keeping a close eye on market advancements, potentially fostering collaborations. Notably, they've already partnered with a firm specializing in point cloud processing add-ons. We're eager to witness the ongoing evolution of this landscape.
The importance of AI integration in the Scan-to-BIM workflow
For many businesses, including ours, it's essential to ensure that clients receive the highest quality and customized products, like CAD models, at competitive prices. Achieving this requires us to leverage advanced and efficient tools to enhance our services.
We take on the responsibility of continually evolving and implementing these tools, while also carefully exploring and testing better solutions to make informed decisions about their deployment.
Are you interested in the topic of 3D modeling (BIM) and want to learn more about it or do you have specific questions? We would be happy to have a conversation about it.
Michael Danklmaier
Miviso Co-Founder
Email:Â michael.danklmaier@miviso.com
Tel.:Â +43 664 4563309
*model requirements: Define the requirements for a model. This includes general information about the building, the level of detail of the finished model, documented information of the building, required file format, and office standards, if available.