Pamela López
Learn in the interview with Pamela López about the benefits and challenges of using artificial intelligence in landscape design.
Pamela López is a landscape architect and a master in architectural design (UNAM), specializing in adaptation (SRC, Sweden) and nature-based solutions that enhance the genius loci and ancestral Latin American landscapes. She teaches at La Salle University (CDMX).
What motivated you to enroll in an artificial intelligence course focused on the production of audiovisual elements?
It was truly a boost to know that I’m part of a digital environment that changes every 2.1 seconds. In fact, I couldn’t even formulate, structure, and type this sentence in that amount of time. That’s how quickly AI continues to ‘nourish’ itself with every question or comment we make. What’s interesting is that, within that same idea of speed, it also depends on human reflection processes, and that motivated me even more because I could learn to delegate tasks to AI—tasks that, while not essential, can be completed in seconds to complement our questioning, our passion for processes, and finding solutions (or more questions) in the short, medium, and long term; becoming explorers, somewhat like those naturalist figures of the 19th century.
How do you think these tools can influence architecture and landscape design?
Suggesting the word “influence” seems to me just the tip of the iceberg in the conversation and implementation of this; we definitely need to talk about the bias that metaphorically appears as a shadow of the technology. If we assume that all people studying landscape architecture in Mexico have access to the internet, a computer, and the registration required by each AI platform to use it; secondly, that all these people identify as someone living in an urban environment with environmental variables homologous to a city in central Mexico, alongside socio-cultural variables from a globalized contemporary structure, then we could only reach a few findings. However, if these tools are part of an ethical construct, from how we feed the AI, what data we provide, how we open the “black box,” and from what standpoint we allow it to join our networks of influence, then we would make a systemic leap in thinking, opposed to the hours invested in “producing information.”
Applying this to the case of a landscape architect and an upcoming project: this project must complete its stages from analysis, site visit, and production of information in less than 8 days. From the start, the site visit may leave you with only 6 days for everything else. What processes can I advance with AI support while I reflect? Iterations on problem definition, diversity in trustworthy information sources, experimentation with images, democratization of knowledge regarding the ecosystemic values that embrace us, to name a few. But who makes the plans? Well, we haven’t reached that point yet. There are Land F/X, LandKit, LandDesign, RhinoLands, which are manipulated by humans but do not produce on their own.
What are the most significant ethical challenges you identify in applying artificial intelligence to architecture and landscape design?
A moment ago, I mentioned the part about technological bias, that, and something we discussed in the course with colleagues: the poor self-criticism we have about the data we validate. As a result, children and young people receive this data as facts. For instance: “a garden (by AI definition) must have grass and perimeter vegetation, perhaps some rocks, one or two trees, and no people.” If a population like the one mentioned looks for and finds that, they will assume that context, memory, resources, biodiversity, or even the protective spirit of the place don’t matter, this will be, by validation, “a garden.”
Beyond the commercial representation of ideas, how do you see the role of artificial intelligence in experimental processes and in solving complex problems in landscape design and regional planning?
Yes, it’s precisely about allowing ourselves to experiment and see it as that, a creative tool for testing and hypothesis validation. We must take advantage of its speed in producing results as we interact with its back-and-forth “hallucinations.” It should be seen as a blackboard in a lab where ideas pile up, erasures happen, new calculations emerge, corrections are made, and “eureka” moments occur. I believe that if we see it as the only forum of truths, then we are missing the opportunity to fully engage our cortex.
We have complexity simply because we are agents in a system that embraces emergencies and chaos in its interactions, which adapt along the way. I would like to contribute to the conversation with this question: When will the revolutionary moment come for us to understand that regional landscapes are the seeds of our biodiversity and the identity foundation that supports urban areas? If that moment is now, then AI could provide a network exchange where we push it to specify worthy alternatives for regional landscapes. For example, I’m thinking about how Geographic Information Systems (GIS) could become automatic and fast, combining data layers from various sources and adapting them. This would make the production of this data more efficient, allowing us to focus on truly understanding the site by observing it, dialoguing with the community of communities, and perhaps generating empirical, dignified, and real proposals.
Another ethical aspect to consider would be the greenhouse gas (GHG) emissions from the electrical and heat sectors, on which AI depends to function. An example is the spatial or physical capacity required by a data center with thousands of machines boxed in, absorbing and emitting heat. Just ask yourself, how many watts per hour (Wh) does a simple question to an AI processor require?
Do you think artificial intelligence can challenge our preconceptions and help us find innovative solutions in landscape architecture? In what way?
Innovation is a huge topic, it seems like we innovate as soon as we start asking questions and daring to change the paradigm without fear of testing along the way. In doing so, we are already laying the foundation for challenging ourselves as landscape architects.
For landscape architects, I believe it challenges us to connect more with the sensitivity of learning through time, with the justice that comes from revisiting the past, present, and future of any agent within the ecosystem. AI challenges us to distinguish between what is human and what is an algorithm, and it pushes us to see the magic in using more boots, glasses, and sketches, versus, as I see around me now, ergonomic chairs, glasses, and imprecise lines on a screen.
In the field of landscape architecture, we must embrace the plurality of those bridges. For now, I’ll leave these possibilities for experimentation:
- **Perplexity**, to document yourself with reliable sources during a site analysis.
- **Midjourney**, to take your sketches further and transform them into edited images.
- **Huggingface**, **Leonardo**, or **Control Net**, to test a text that turns into an image during the preliminary design stage.
- **Krea AI**, to generate real-time experiments during the conceptual stage.
- **ECOGEN**, a prototype SWA is testing to incorporate into **RHINO – GRASSHOPPER** and finally let those plans almost “make themselves.”
Photography: All images are processed in co-authorship with the Stable Diffusion 3 Medium AI from Hugging Face, using prompts in Spanish and the following default parameters: guidance step (5), number of inference steps (28), avoid negative prompt.