Artificial Intelligence in Architecture: Balancing Efficiency Gains and Environmental Costs

Imagine having an idea for a new architectural project, and instead of drawing or modeling it on the computer, simply dictating your idea through audio and having the project ready in moments. "Hey Siri, create an apartment floor plan with three bedrooms, two bathrooms, an open kitchen connected to the living room, and a terrace facing the side with the best city view"; and within seconds, you have a detailed plan with all the correct fixtures and dimensions.

This is what Artificial Intelligence platforms will provide for us designers and architects.

OpenAI's GPT is one of the platforms that promises to change the game for creating elements without requiring any programming knowledge. With more than 175 billion parameters, the platform can generate scripts, write articles, create works of art, and develop applications, using only voice commands and without additional human supervision.

By searching for patterns in its database, the platform suggests solutions that fit the parameters of the audio recorded by the user.

GPT stands for Generative Pre-training Transformer, which uses Machine Learning algorithms to automatically learn from its experience without needing to be reprogrammed.

For Architecture, this means building a database with thousands of years of experience and experiments that can serve as a repertoire to create a new kind of Architecture that is informed, customized, and connected.

We can no longer deny that technology is rooted in Architecture, whether through CAD, BIM, Parametric Tools, or more recently, Generative Design. As Architects, we must master these tools to generate forms and spaces that enhance life on earth (and beyond).

Artificial Intelligence will no longer be about defining only the optimal orientation of a room, or its geometry, thickness, or texture, but about expanding the consideration to include context, use, history, user type, and local materiality, generating design options that are highly informed by precedents and optimized typologies, becoming not only an extension of the pencil in our hands but of our senses and intelligence.

Tools like GPT-3 can help streamline the creation of multiple variations of a single architectural design, whether based on a style or on a specific organization. Just like generative models, Artificial Intelligence analyzes the repertoire database and the requirements imposed by the user, and qualifies them, ranking the possibilities from the least to the most adapted.

These tools become part of the design process, with a holistic view of architecture, extracting not only quantities but also qualities, understanding and dissecting all the factors that influence the process of spatial development through trial and error.

On the other hand, we must be aware of the environmental impact that this technological power brings. Storing 45 terabytes of memory requires intense energy consumption, significantly increasing the carbon footprint. Studies indicate that training an Artificial Intelligence processor can generate more than 280 tons of carbon dioxide, equivalent to five times the emissions of a single car.

Knowing this, we must ask ourselves how to develop new tools that advance the practice of architecture but use technological intelligence in the most efficient way possible, benefiting the user without negatively impacting the environment. Renewable and sustainable energy must be at the core of the discussion, driving intensive investment in research and the application of new alternatives, carefully weighing the balance between the gains for architectural design and the environmental cost.