Do construction robots dream of electric sheets? And other questions for the future of AEC

Alex Serriere
February 8, 2023
Min Read
Do construction robots dream of electric sheets? And other questions for the future of AEC

This article was written by Alex Serriere, Principal, Executive Vice President, located in San Francisco, California. Alex is TEECOM's Chief Technology Officer (CTO) and Director of R&D, keeping our team and clients up-to-date with the latest technologies. For over 15 years, Alex’s research and leadership has enabled TEECOM to plan for future trends and insights and maximize long-term flexibility. Alex is driven by the notion that advancements in technology bring something new every day. Today, Alex’s team focuses on developing new tools for designing the world’s most complex facilities. Tomorrow? We have research for that!

AI-generated image of a robot vacuum on a construction site in the foreground with a building in the background and a sky made of blueprints.

Replicants may still only be a feature of science fiction futures, but in the present, numerous organizations are experimenting with robotic systems to accelerate construction projects of all kinds. As these systems mature and advance, it’s worth considering the role of construction deliverables in a future where more and more tasks are performed by robots.

Humans have always been the primary audience for construction documents such as specifications and drawings, but large scale 3D printing, autonomous rebar tying, and robotic drywall finishing aren’t driven by that kind of documentation. And while we’ve seen a few use cases where portions of a building information model can be extracted and fed into computer aided manufacturing systems without too much trouble, these initial steps into robotic construction still depend on humans to program them with information derived from legacy deliverables.

If we envision a future where robots are more than a curiosity at construction sites, how will we move beyond traditional construction documents that are designed for human consumption? What do the specialized formats for robots look like? Is there an AEC-specific G-code that enables machines to read and interpret this information directly instead of passing through a lossy series of digital construction document sheets?

This is not to say that human-readable deliverables will become obsolete in a future dominated by robots. However, it's clear that the way we create and use these documents will have to adapt to the changing nature of the industry. As we move towards a more automated future, we should consider the needs and capabilities of the technologies that will be performing the work, adjust our processes accordingly, and not continue doing things just because that’s how they’ve always been done.

Does the metaverse need security cameras?

If the metaverse happens (and that’s a big if) it’s easy to imagine that these virtual worlds will come with a wide range of built-in security mechanisms that prevent nefarious actors from, well, being nefarious. But it’s equally easy to imagine humanity’s apparently unbounded ingenuity when it comes to mischief, harassment, and generally breaking the rules. So, like any online community, that means the metaverse will need codes of conduct, reporting mechanisms, and moderation, but unlike text-based communities or communities built around recorded video, does the dynamic nature of metaverse environments mean that public spaces will also need mechanisms for capturing and playing back suspicious activities?

AI-generated image of security camera footage from the metaverse with 3d avatars.

Maybe that recording mechanism will be built-into every metaverse client, maybe each metaverse will, ultimately, have all the necessary compute and storage capabilities to record every interaction in perfect fidelity, but the properties recordings created by security camera are straightforward and well-understood. Predictable perspective, authenticated timestamps, and configurable retention periods, these are all valuable features that could be transposed into virtual environments. And in the same way that we can get licenses for virtual appliances for various networking equipment today, perhaps camera manufacturers will begin offering licenses for virtual cameras deployed in the metaverse. That way, recordings can be integrated into existing Network Video Recorder (NVR) infrastructure and record-keeping processes.

And while we’re thinking about security cameras in the metaverse, perhaps we should consider applying Crime Prevention Through Environmental Design (CPTED) principles to our virtual environments as well.

Does an AI architect need a license?

The rapid advancement of tools like ChatGPT, DALL·E, Stable Diffusion and more have demonstrated that all kinds of professions will eventually be impacted by artificial intelligence. The AEC industry, while often slow to adopt cutting edge technology, will undoubtedly be impacted as well.

It’s easy to see how a first pass at specification documents or placement of various components on a floor plan could lend themselves to the sort of probabilistic knowledge contained within machine learning models. In fact, asking ChatGPT to write specifications today produces output that, at first glance, is entirely reasonable. Indeed, it’s not much of a leap to see how something like DALL·E’s Inpainting could be used to rapidly populate a massing model for a new facility — from sketches to construction documents in mere moments.

AI-generated image of a robot doing architecture work in the style of blueprints.

As practitioners become more comfortable and more reliant on these tools, it prompts  a question of accreditation and licensure. Unlike a traditional computer-aided design (CAD) application, these tools have the potential to begin making decisions on their own. Not in a sentient way of course, but in the same way that we have begun asking autonomous cars to make decisions about driving. We can say that the driver is ultimately responsible no matter what, but that doesn’t change the fact that sometimes these computer systems will make the wrong decision. That raises the question: do we need some sort of licensure system to ensure baseline capabilities in AI systems applied to AEC? 

With cars, many states have passed legislation requiring manufacturers to acquire permits for testing and deploying autonomous vehicles. These permits also require a certain amount of reporting from the manufacturers to ensure the safety of these systems as they interact with the public. Perhaps state architecture and engineering bodies should undertake this kind of initiative in the near future as well.

If you’re as excited about harnessing these tools and advancements as we are, whether that’s to design better buildings or to improve the design process itself, drop us a note to learn more about how our R&D group enables us to better serve our clients.