Archinect's Lexicon focuses on newly invented or adopted vocabulary within the architectural community. For this installment, we're featuring an artificial intelligence term that has been cited in our recent feature articles with both Matias Del Campo and Genevieve Goffman.

A "Generative Adversarial Network" (GAN) is a class of machine learning systems invented by Ian Goodfellow and his colleagues in 2014. GANs consist of two parts, a generator and a discriminator, that are set against each other (thus the term 'adversarial').

The generator creates new data instances, while the discriminator evaluates them for authenticity; i.e., it determines whether each instance of data that it reviews belongs to the actual training dataset or not. The generator tries to fool the discriminator, and there is a continual back-and-forth as each side learns from its adversary. This eventually enables the generator to produce high-quality data that are almost indistinguishable from the real ones, as judged by the discriminator.

GANs are relevant to art and architecture because they can be used to generate new examples of art or design, synthesize elements within an image, or even invent entirely new forms of visual expression. They can learn from large databases of existing works and then produce their own designs that mimic the style, form, or content of these works, often with uncanny results.

Architects can use GANs to generate new design ideas. Given a dataset of building designs, a GAN could generate new architectural structures. This could help architects to explore new forms and design possibilities, or to generate a large number of design possibilities quickly, which can then be assessed and refined by human designers.

Moreover, GANs could be used to create detailed and realistic renderings of architectural designs or to synthesize elements within a design such as furniture or fixtures. They can be used to experiment with different design aesthetics and to imagine how different design elements might come together within a space.

However, it's also important to note that while GANs can produce amazing results, they are merely tools and the quality of output depends on how they're used. Furthermore, they don't replace the creativity, knowledge, and judgment of human artists and architects. They also raise important questions about authorship, creativity, and the role of AI in creative fields.

This article is part of the Archinect In-Depth: Artificial Intelligence series.