Claudia —
Over the last two decades, your research and practice have been crucial for design and architecture in redefining robotic fabrication. How would you describe the role of machines in your work? On a spectrum with simple tool at one end and intelligent collaborator on the other, where would you place machines?

Kohler —
I think we look at the robot, in itself, more as a simple device, with motors and without any intelligence, that acts as a bridging device, rather than as a traditional tool. Conceptually, it enables this technical bridge between two domains, between the physical and the digital, which brings the question back to us: What do we want from this bridge? 

That’s why we chose to start with industrial robots in our initial work during the early 2000s. Since they don’t have hands, you need to think about what they make. In that sense, it is a trigger for human intelligence, pushing us to define digital craftsmanship. Practically, it is a customised use of industrial robots within additive fabrication processes, but it also enables us to explore the interplay between digital control and physical action.

At the same time, the robot’s limitations create a different kind of intelligence. Unlike a tool of unbounded possibilities, such as artificial intelligence, the robot is constrained, so it triggers the human intelligence of what to do elegantly within this limited set of possibilities. Interestingly enough, this opens a new design space, which appears through the limitations rather than through an intelligence of the machine. Of course, that was 15-20 years ago. Now, with humanoid robots, things are shifting, but that’s basically where we come from.

Robotic experiments at the MIT Artificial Intelligence Laboratory, 1968. Copyright: MIT History Museum

Gramazio —
I think it's a very good definition. To extend it slightly, the tool doesn't have to be intelligent in itself, it can also be a media. Tools, like a screwdriver, or photography are technical developments that don't really think like humans do. But it is not only about human intelligence; it is also about considering design as a process, where intelligence emerges through dependencies.

As Matthias said, without the possibilities and also the limitations, design intelligence doesn’t develop. The tool itself may be simple, but both possibilities and limitations drive intelligence.

Tom —
When you mentioned new design spaces, I immediately thought of Nicholas Negroponte, who, in 1975, envisioned architecture as an adaptive and interactive system in Soft Architecture Machines.  He had the idea of public-machine collaboration which would shape built environments dynamically, ultimately replacing the architect.

Now, if we use a "dumb" robot or tool to create a new design space, I imagine a future where anyone can write a prompt on their phone. A robot then arrives in a modified freight container and erects the building. You described something similar in “R-O-B”, your 2008 mobile fabrication project. Would you say that this is a possible future? Could something like this happen in the next 20 to 50 years?

Kohler —
Yes, technically, that’s absolutely possible, but it's more a question of, where will it be? Who will want this? Why would people want it?

Technically there are not so many limitations. We explored mass customization early on, like the m-table, which was a customizable table series from 2002, where you could design a table on your phone, and it could be directly manufactured. But the real question was never Can we do it? It was, rather, Who wants it?

Imagine prompting a robot to build something, then realizing that the result isn't something that you wanted. How do you correct it? The complexity of the system, the risks, and the extremely high cost, make people hesitate.

The challenge isn’t technical; it’s about market demand, user acceptance, and cultural expectations. It’s not as simple as ordering a Tesla in a different color.

Gramazio —
And I would also be careful with the "prompt-to-design" idea. It might become a design methodology, but not a tool of unlimited possibilities.

As far as I remember, Negroponte wasn’t talking about substituting; he was pointing out that using computers just as drafting machines was lacking in meaning. Now, since machines are becoming more intelligent, his vision becomes more plausible. He imagined the machine as the computer, which becomes a partner to the designer and engages a dialogue on eye level. While there are still capacities and skills the human has that Chat-GPT doesn't have, prompting becomes interesting for us as a design tool.

Of course, it's possible that the industry misuses artificial intelligence to bypass the architect. If we are realistic, less than one percent of buildings worldwide are designed by architects. So why not have Chat-GPT design buildings? The profession and the notion of design can profit from new methodologies, but only if we proactively embrace these new developments.

RIBB3D, ETH Zurich, 2021–2022. Rethinking concrete floors through 3D-printed formwork—this research explores material-efficient ribbed slabs that reduce CO₂ emissions by up to 40% without compromising performance. Copyright: Gramazio Kohler Research, ETH Zurich and TUM Professorship of Digital Fabrication

Claudia —
What if we think of machines differently, not just as tools for production, but for participatory design? What potential do you see?

Gramazio —
In that sense, it has a lot of potential, because it takes a role in the more complex landscape of the design process, along with different technologies, diverse people’s needs, various stakeholders, multiple tools to process materials, a culture around us, an availability of energy, etc. Design has never been about one genius making a sketch. It has always been more complex. Intelligent tools could help us rethink the process, making it more collaborative and moving beyond the myth of the genius architect.

Kohler —
Thinking of co-design, the application of machine intelligence as a mediator could be meaningful. But digital technology already plays this role, if we look at the internet’s platform logic which enables a collaborative system.

But as a slightly critical thought, I wonder if there might be a tendency to delegate decisions to these systems, because it's comfortable in collaboration. Negotiation can be difficult. Interests often conflict, and it’s tempting to let a machine arbitrate, which basically takes decisions away from you.

Imagine a virtual judge processing building laws that tells you if a design is permitted. Authorities could create it, and designers have the benefit to constantly check their designs against it. It would be efficient. What this does in the end, though, is  offload our responsibility for the judgment to the intelligence.

And while I wonder where this leads, the convenience to upload everything doesn’t only lessen your responsibility and autonomy. It also creates a lack of critical debate and an active dependency on intelligence. This less creative use of artificial intelligence is basically a self-decided submission under the intelligence, which could be misused by certain people in power.

Tom —
Nowadays, we are faced with new tools and systems every day. Especially in architecture, the roles and skills required are quickly evolving. Which new skills should architects focus on, and how do you teach them?

COMPAS XR, 2024–2025, An open-source tool for architectural research and education, streamlining XR workflows for real-time collaboration and spatial visualization. Copyright: Gramazio Kohler Research, ETH Zurich and TUM Professorship of Digital Fabrication

Gramazio —
Certain practices, like hand-drawing plans, have become obsolete, and new tools keep emerging. Most schools, including ETH, are figuring out how to integrate this as something more than just extra knowledge. It's important to learn but as a fundamental part of the profession.

However, practical skills, like using digital modelling tools or BIM, can be picked up along the way as the industry evolves. More importantly, I think the concept of the agency is crucial, both in technology and society. Over the last few decades, architects have stepped back from being central figures in decision-making and have been replaced by specialists.

As a result, architecture has become somehow marginal to those investing in cities and buildings. It is a question of attitude; schools should teach architects to take responsibility beyond design, to engage with technologies, materials, resources and societal needs. Once students grasp this concept, they independently seek deeper knowledge pushed by the competence and passion that are key to becoming a complete designer. This is why we explored robotics—not because we had to, but out of curiosity about how it might change architecture. That sense of curiosity and agency should be at the core of the profession.

Claudia —
That’s really interesting, especially the idea that curiosity drives innovation way more than obligation.

Tom —
And it's refreshing to hear a perspective driven by curiosity, in contrast with how easy it is to feel pessimistic about the future, especially in architecture, with today's challenges.

Gramazio —
Absolutely. Teaching today, we see how overwhelming global issues like climate change and geopolitics can feel. But I always recognise that my students, compared to students of our generation, have a greater opportunity to make architecture truly meaningful. When Kohler and I studied in the ‘90s, the architectural discussion was mainly focused on form and abstract ideas. We were influenced by late-modernist and early-postmodernist thinking. The pillars of architectural discussions were purity of shape, structural expression, and formal symbolism. Now, architecture has the immense potential to engage with real-world, urgent problems. The key is in the shift from a sense of despair to proactive engagement. It completely changes the perspective. It's no longer just about form; it's about making a tangible difference.

Claudia —
On a different note, since you work in parallel with machine and human intelligence, we are very curious to know how you define intelligence.

Kohler —
Two words come immediately to my mind: critical thinking and creativity. Critical thinking involves reflection and even deliberate disruption, whereas creativity allows for re-framing and intuition. I think they're closely linked: one challenges, the other reimagines.

Artificial intelligence can do both, but without human interaction it remains limited by the pre-defined datasets. Of course, one could argue that humans are also shaped by prior knowledge, but that's part of an ongoing philosophical debate. Ultimately, we still don't fully understand intelligence, both human and artificial, which is exactly what makes this conversation so fascinating.

Gramazio —
I agree with your definition. Intelligence isn't a fixed concept. We have underestimated what machines can do repeatedly throughout history. The assumption that they can only learn from existing data might be flawed—after all, humans also build on existing knowledge but have still managed to innovate. Maybe we're asking the wrong questions. Maybe it's not so much about the definition of intelligence. Maybe the future of intelligence, whether human or artificial, lies in areas that we haven’t yet defined.

Marvin L. Minsky with his computer-controlled robot in the MIT Artificial Intelligence Laboratory, 1968. Copyright: MIT History Museum

Tom —
Let's wrap this up with one final question. You have been pioneers in robotic fabrications. What are the biggest challenges that you are facing now?

Kohler —
That's a good question. One of our major challenges is that architecture and construction tend to remain highly risk averse. While, on the surface, things seem to be changing quickly, deep-rooted traditions resist transformation harshly. For instance, we've explored AI-assisted design and built projects like the semi-automated vertical garden in the Tech Cluster in Zug, which combines computational design and robotic fabrication to create a plantable architectural structure that allows nature to grow undisturbed. In-house this works well, but the moment we engage external partners, we always encounter hesitation.

In our immersive design studio at ETH, we have eliminated plans, working instead entirely with 3D models and VR. The most common reaction we get from external people is that they feel as if architecture itself is being lost. This deep-rooted resistance is one of the biggest hurdles in adopting new methods. Even after 20 years of pushing innovation, changing construction is still slow and much of what we developed feels like a tiny drop in the ocean.

Tom —
This has been such an inspiring discussion. It's reassuring to hear that while change is slow, we still have the power to shape the future of architecture, which is very much an empowering perspective on the overwhelming challenges the world is facing nowadays.

Claudia —
Absolutely. It's fascinating to see how different approaches, like eliminating plans, can completely shift our perspective on the profession. We often assume that we're up to date as students, but there's always more to explore.

Kohler —
Yes, and it looks like you're engaging with important topics already. That’s great to see. Keep challenging your perspectives!