MVRDV - Leo Stuckardt of MVRDV NEXT discusses data, AI, the Metaverse, and more

Leo Stuckardt of MVRDV NEXT discusses data, AI, the Metaverse, and more


MVRDV has long been known for its affinity for technology and data, from its early embrace of digital visualisations and creation of "datascapes" such as MetaCity/DataTown, to its ongoing development of custom software for both internal, project-based use and sharing ideas and concepts with the public. In recent years, this focus has been formalised with the creation of MVRDV NEXT (standing for New EXperimental Technologies), an expert taskforce that implements advanced technologies – both within the firm's broader portfolio, and as standalone projects. In this extensive interview Leo Stuckardt, a senior project leader for MVRDV NEXT, discusses the current technological landscape within the field of architecture, from the need for accessible urban data to topical subjects such as artificial intelligence and the metaverse. A shortened version of this interview was first published in Chinese in VISION Magazine.

The shape of Valley's complex arrangement of terraces was heavily dependent on scripts developed in Grasshopper to optimise considerations such as daylighting and sightlines. Image © Ossip van Duivenbode

MVRDV’s research covers multiple fields such as architecture, urban planning, sustainability, sociology, material development and more. What kind of digital technology, artificial intelligence, or anything else will help you make better decisions?

First and foremost, I think what is needed is open data – primarily urban data, but also environmental and data on materials – and technologies to access this data, to contribute to it, to share it, to verify it, and to visualise it. This is crucial, in the sense that we need to develop a deeper understanding of the contexts we engage in. Secondly, our projects are increasingly steered by environmental simulations and assessments. We’re quite advanced in this already and forecast the impact of all our projects with regards to energy requirements but also to evaluate spatial qualities for inhabitants and the biodiversity that surrounds our projects – so really all the users of future developments.

Beyond that, of course, we employ parametric design in various ways, such as scripting within Rhinoceros and Grasshopper to quickly explore options, to rationalise designs and make things more efficient. We also built our own tool, CarbonScape, which we use to assess carbon footprints or the sustainability performance of buildings.

On the experimental side, we are exploring the use of game engines. So that includes Unity and Unreal Engine for visualisation, but also for interactive decision-making processes. That way we can involve communities, clients, or political stakeholders in a playful manner. At times, this can help us to go beyond the constraints of a classical design brief and test “what-if” scenarios that might challenge assumptions that would otherwise go unquestioned.

Now, of course, artificial neural networks or other forms of machine intelligence are increasingly becoming a key factor and I think they will challenge the way we work. At this point, for the most part we can only speculate on their impact or how they will be applied to architecture. Needless to say, we are following developments here carefully.

Have you got any specific examples you can give in terms of those applications?

Generally, I believe that there will be several applications. Maybe the most foreseeable at this point is the development of so-called “surrogate models”, which essentially replace otherwise-costly simulations or calculations. They do this through neural networks that can approximate the results of these calculations – but instead of taking three hours to calculate, they might do so in a few seconds. While the result of a single calculation might not differ by a lot, their potential impact is enormous as this allows to test significantly more variations in a fraction of the time.

We also use machine learning to explore and map our own digital archive, to try to filter and sort through the terabytes and terabytes of data that have been produced at MVRDV. Imagine how all this material, from construction details to bold, unrealised visions could be re-used if only we could access the right information at the right time.

So in a nutshell we expect a strong impact of AI models on both, the generative side as well as the analytical tasks of the design process.

Developed as part of the MVRDVHNI exhibition in 2021, MVRDV Cloud uses computer vision and metadata analytics to map the contents of MVRDV's digital archive to produce new narratives and support the development of future projects by MVRDV.

So, with the help of new technology is the work of modern architects easier or more complex than it used to be?

When looking at new technologies through the lens of digital planning and design tools I would say it's neither. Design workflows are changing rapidly and at the same time, the workload stays the same. Automation and digital tools of course speed up production. We can create a lot more content and test options much more quickly. But we also notice that the expectations of clients increase. For instance, ten years ago a competition booklet at MVRDV might have been between 100 and 200 pages. Today, similar booklets can grow up to as much as 1,000 pages. So we produce faster, but there is also more expected. That's one part.

The other, more critical part, is an increasing challenge of technology literacy across generations of architects. Technological developments accelerate the means of architectural production constantly. The software we use to draw, model, calculate, but also communicate projects change fundamentally within a single generation of architects. The tools that we learn during our academic education will become obsolete throughout our career. This was not the case in the previous century and requires us to constantly learn, to stay up-to-date, and even to educate our colleagues and collaborators.

Facing that development of artificial intelligence and this speeding up of the pace, as you mention, what can architects do?

Particularly when looking at artificial intelligence, it seems these recent developments are shifting towards centralised, proprietary models. Large companies such as OpenAI (GPT 4) own models that cost millions of Euros to train and to develop. Hardly anyone else has the ability or the resources to create AI models of similar capacities. So we risk a disparity with regards to authoring and deploying large, artificial neural networks.

I think there are two things that architects can do at this point. Firstly, engage with these emergent models critically. At this point we are probably all aware of discussions on AI bias within social, cultural or legal contexts, but where are the blind spots of these models for inclusive design processes for buildings and cities?

In addition to that, I believe architects should put more effort into archiving their work responsibly and with future use in mind, whether that will be the training of AIs or other forms of accessing information for creative reuse. I think that these repositories – the servers of architectural companies – will increasingly be relevant to prevent a homogenisation of architecture through standardised, white-box tools.

Generative AI techniques allow for the rapid visualisation of ideas through images. For this strategic project, an interactive toolbox was linked to Stable Diffusion to visualise the potential consequences of an interactive, participatory design decision-making process

In the design and construction of Valley the experts at MVRDV NEXT created a series of custom digital tools to perfect the building. Can these digital tools be used widely in other projects?

At MVRDV NEXT, most computational tools are developed to be highly project-specific. We put effort into modular approaches, which allow us to reuse or share parts of these solutions but the overall workflow is usually tailored directly to the needs of a project. In part this is due to factors such as the location of a project or specific collaboration models with engineers, clients or other parties involved. This balance between unique, custom software and scalable solution compliments MVRDV’s overall design method well in my opinion. Every project is part of an overarching vision, while being a highly unique prototype.

With regards to Valley, the main scope of work centred around the rationalisation of the façade. It would be hard to transfer that one-to-one into another project because we would work with different façade engineers, manufacturers, and of course a different architectural concept.

In addition to that, these technologies age quite rapidly. In most cases it takes years to complete a project, while technologies develop a lot faster. So while we don’t reuse the complete tool in most cases, we can still reuse parts of the code or the methodology. We might then build a new computational tool that inherits some functionality or ideas, while giving it more up-to-date technologies and capabilities.

Valley also required a complex scripting process to optimise the layout and installation process of the seemingly random pattern of stone tiles.

After rationalisation and design optioneering, what is the next experimental technology of NEXT?

Of course, there is the aforementioned exploration of machine learning or AI for analytical and generative applications.

At this moment though, we’re very interested in technologies for storytelling – exploring the role of video, gaming, sound and interaction to convey the main ambitions or challenges of making architecture. We notice that, as our projects become increasingly process-oriented, we need to become analysts as much as we are creators. What kind of languages and narratives are needed to convey decision processes that are based on wide sets of numbers, instead of the sole shape of a building?

In addition to computational designers we now have a growing number of specialists for carbon and life cycle assessments, experts for biodiversity strategies, and experts for sustainable construction techniques. Their role is not only to consult ongoing projects, but also to collect data and to evaluate the overall success of strategies that were employed within projects. So we’re curious about how we can collect this information, tell those stories, and use conclusions as tools to negotiate briefs with clients or raise their ambitions towards ecology, sustainability, and so on.

In addition to that, one of our focuses is the use of interactive models. We are enthusiastic about game engines, since more immersive, reactive virtual experiences can also help with this storytelling. In my opinion, designing a shift from static digital models towards dynamic, responsive ones is the most exciting, and one of the most crucial developments in architectural tooling. Imagine models that push back, prevent or suggest certain design compositions according to forecasted impact.

What do you think is the importance of that approach to analysis and storytelling for NEXT?

The discussion in the making of architecture is becoming less and less output-oriented and more process-oriented. We see this in the increasing number of consultants that are being taken on board, but also in the role of MVRDV in architectural and urban design projects. It appears to be more and more a role where conversation, feedback groups and these kinds of things are the central task. It's hard, then, to centre that discussion around the section or the physical model of a project. You need to centre it around the process: how were decisions taken, which priorities were given, under whose influence… In a way, architectural projects are stories more than objects. We can no longer convey these stories through drawings alone, but oceans of pictograms and icons might be even less helpful.

The RoofScape software makes use of urban data to allow users to investigate scenarios for the uses of a city's rooftops.

Could you talk about how you started the RoofScape project?

We could really go back to the beginnings of MVRDV. Density has always been a fascination of MVRDV. It was kind of a natural match to look up and to discover these empty spots on top of our buildings. Rooftops are simply underused in most European cities. We noticed that while density is not the single answer to urban challenges, it can significantly contribute to social, climatic, and sustainable urgencies. Activating a vacant rooftop and putting additional housing there, if done in the right way, will create affordable space and help with the housing crisis that we currently face in most European cities. Putting green rooftops there will help with climatic issues such as urban heat islands or floor risks. Placing photovoltaic panels on top of them will increase energy production. So, we see a wide range of potentials there.

The next step was MVRDV’s investigation of rooftops through local, temporary activations, such as the Rotterdam Rooftop Walk and The Podium last year, but also the Stairs to Kriterion in 2016. After that MVRDV published ‘The Rooftop Catalogue’, which collects a series of qualitative imaginaries. It asked, what kinds of spaces can we create on those rooftops? What could they look like? Why would we desire to be there?

We now had a lot of material that conveys the dreams and proves the spatial qualities. But we felt that we needed to add the quantitative, systematic side. What if we zoom out and look at the big picture of the city, what is the potential of scaling this up?

So we built an interactive model in the game engine Unity: The RoofScape application measures potential impacts of rooftop activations – how many megawatt hours of energy can be generated, how many square kilometres of parks can fit on the rooftops, how many new apartments can be constructed? At the same time, RoofScape allows us to simulate scenarios on the interplay between these different programmes.

It’s important to emphasise that we didn’t design this with the vision of a singular “ideal” city, but in order to understand the influence of different steering mechanisms or public policies. What happens if we stimulate or incentivise the activation of green rooftops in an area? What form of habitable spaces would it create? Could the quantitative benefits of it help finance construction efforts?

Projects like RoofScape expand on MVRDV's work regarding rooftops, with interventions such as the temporary Rotterdam Rooftop Walk of 2022. Image © Ossip van Duivenbode.

In your opinion, what kind of architecture or cities do we need in the future? 

I would say diverse architectures and heterogenous cities that allow for a multiplicity of dreams, actions and ways in which spaces are created. Obviously this has to include the desires and needs of animals and plants to create a healthy environment.

In addition to that, we need cities that are inhabited. The financialisation of the city is one of our biggest challenges at this point: From real-estate speculation that results in vacant apartments up to entire ghost towns, we see buildings becoming financial assets rather than places to live in. At times, as architects we might feel powerless to engage with these processes but we're trying to figure out what our role and the role of spatial design is in this.

If more and more people flock to big cities, how can we achieve the balance between city and nature?

There are two possible perspectives towards this question: what to do about the city, and what to do about the spaces in between cities. We know at this point that the cities themselves need to be contained to avoid urban sprawl. So part of the challenge has to involve creating density in (ecologically) responsible ways and to keep cities compact. At the same time, we need to bring in greenery and biodiversity to create healthy, resilient, cool environments.

In contrast to that, we might need larger areas without human interference around the cities as well. With the ‘Half Earth Theory’ Owen Wilson proposes a fascinating model of the planet, were half of it is acted on by humans, and half of it is completely left alone. Complete wilderness, no agriculture, no logistics centres, untouched. The theory claims that if we were to follow this division most crises that we are facing now could be avoided. How could we design this process of retreat?

Technology will continuously challenge our design. What kind of architecture will have more resilience to environmental change?

At this point, choosing refurbishments over new construction whenever possible is crucial. In the Chinese context MVRDV has pursued this paradigm with the Idea Factory and with Shenzhen Women & Children’s Centre. That is a large part of it. The aspect of change, the idea of flexible design, extending the lifespan of a building by making it attractive and adaptable to change, has to be paramount both for how to engage with existing building stock as well as with how to conceive new constructions if they can't be avoided.

Contributing to this of course will be the use of renewable materials. Bio-based materials, timber, or earth wherever possible. In addition to that we need to add strategies for demountability in buildings or parts of buildings, modular design approaches, and the integration of landscape, vegetation, irrigated surfaces to avoid urban risks such as flooding or urban heat islands.

Matrix ONE is a building that is demountable, with over 90% of its materials being reusable in future applications. Over 120,000 components of the building are visualised here using information extracted from the project's Revit model. 

To bring that answer back to the topic of digital technologies – historically, computer models and simulations have favoured a tabula rasa approach. There used to be a sense that the existing world is too complicated and introduces too many variables. Now you’re working more with the existing real world, like with RoofScape where you take the city as it is and build on it as part of the simulation. How does that change the way that you work?

Looking within the computational domain, there certainly is a possibility that models can be transferred from a tabula rasa into more complex environments that take the contradictions and particularities of existing places into account. It's easier to test and verify them in a sort of neutral petri dish and then increase the complexity later. But in many regards we also realise the limitations of transferring tabula rasa models into specific contexts. Technological understanding, and in consequence data-driven models of cities, is deeply embedded in local cultures, specific urgencies, specific terminologies, or specific decision-making processes. So how could you design a digital tool for a city without trying to understand that very city first? We are now in the beginning of expanding the RoofScape application from Rotterdam to cities beyond the Netherlands and already realise that large parts of the methodology, but also the visual language of the application will need to be adapted.

A second reason for the shift away from theoretical – or tabula rasa – models is the rapidly increasing amount of data that is being collected. From satellites to CCTV, open street maps, and weather stations, cities are constructing their own sensory apparatuses. The availability of this data allows designers to build specific technologies for these places. This is why I prioritised open data in the first question over popular technologies such as AI, VR, or 3D printing: having information about places – large amounts of diverse sets of information to pick from in the right way for the right purposes is a foundation for all design technologies to come.

This holds especially true for a globally operating practice such as MVRDV. Think about even the impact of Google Earth alone. It helps us to work responsibly in contexts that we're less familiar with, or that we have less access to.

What do you think of metaverse architecture? Is it a new territory for architects?

It's definitely an interesting territory, although I don't know if it's a new territory! If you look at the history of computer games, or the often-quoted rise and fall of Second Life, the recent hype around the metaverse builds on a lot. Architects have been fascinated about the virtual domain for a long time. I'm thinking of Asymptote’s Advanced Trading Floor for the New York Stock Exchange or MVRDV’s Metacity/Datatown, for instance. Similarly to AI models, it is not the idea of the Metaverse that is new but rather its accessibility through powerful computers, mobile phones and AR/VR technologies that come with it.

To be fair, it is a bit disappointing to see what has been offered by architects in recent times. For the most part the Metaverse has been a place to revive discarded design options in virtual space or speculate on NFT based, virtual real-estate. But there are real, interesting questions to be posed. How could we work in a world where material resources need to be considered less than polygon counts? How could we work in a world with zero gravity?

Most importantly though, we need to overcome the idea of the Metaverse as this singular, clean, Euclidean space, in which people place assets. Similar to our physical environment or even the internet, the Metaverse probably spans across many proprietary and open platforms. Maybe Microsoft Teams is part of the same, heterogenous Metaverse as Fortnite or Minecraft? How do we then want to move between these spaces? Designing these moments of transitions and connecting different technologies seems like a more interesting question to me. What would the spatial version of the Google search bar look like today? I would hope that architects can have some influence on it.

Developed in 2020, SolarScape is a virtual computational model that visualises Rotterdam's policies regarding sunlight. It could be considered a form of "Metaverse", in the sense that it is a digital rendering of space that allows policy-makers, planners, and citizens to negotiate the future of their city. 

What responsibility should designers and architects take on in the age of artificial intelligence?

As I discussed before, AI has the potential to do things a lot faster and more accurately than us. But it can also confuse and overwhelm. Part of the task is to counter it with human intelligence, with empathy and with – I guess with a form of reduction, keeping focus on certain things. At times it feels we're being overloaded with information and digital content. At best, this confuses and misleads design processes. At worst, it manipulates them. So, designing with empathy might be a good keyword.

Besides architecture, what issues are you concerned about?

Of course, the planetary climate crisis concerns us, as we are involved in one of its most harmful industries, while feeling increasingly powerless. I also think increasing economic inequalities have to be a key concern, for everyone personally and for architects in particular. As architects we are torn between being dependent on and enabling financial speculation and aspiring to design for people, animals and the planet. At MVRDV we try to address these kinds of concerns beyond what can be achieved in commissioned projects. We initiate research projects, conversations or speculative design projects. We participate in public events and conduct workshops and design studios at universities to share our insights and learn from others.