GHASSAN ALSERAYHI
 MSc. Arch, M. Arch, B. Arch, Assoc. SCE

ARCHITECT + EDUCATOR + RESEARCHER
Computational Evolution and Learning:
Data, Subjectivity, and Authorship
Unraveling the Complexities of Machine Learning in Design


  In the realm of architecture, Jenny Sabin's "Embedded Architecture" and Nicholas Negroponte's "Soft Architecture Machines" texts explore distinct yet interconnected paths when it comes to their content's purpose. While Sabin's inquiry prompts critical examination, Negroponte's narrative provides answers. However, both texts converge in their discussion of personalizing space and the intelligent environment. Sabin's Ada project, for instance, exemplifies the direct integration of cutting-edge technologies such as artificial intelligence (AI), advanced geometries, and innovative fabrication techniques. Through these elements, a novel world or microcosm is created, captivating users on an emotional level. This environment operates on the basis of human-data collection, transforming participants into active contributors and operators. The project functions as an infinite loop of engagement, where the utilization of AI constantly involves and unpredictably immerses participants in the project itself. Their mood, emotions, feelings, and movements within the space generate a continuous stream of data that informs the behavior of the knitted exoskeleton, dynamically altering its colored lights with various gradients. Thus, Sabin's Ada project embodies an environment where behaviors are exhibited by the physical surroundings, echoing Negroponte's anticipation of a similar concept.

While Negroponte expresses doubts regarding the underlying objectives of the intelligent environment, Sabin's team aims to expand and inspire human emotional engagement through the exploration of beauty and materiality. In Sabin's and Negroponte's examples, the pivotal role of data, particularly human data, becomes evident, transcending its specific nature. Participants assume the role of community partners, consciously and subconsciously redefining, manipulating, and altering the microenvironment they inhabit. An analogous illustration can be found in the work of the Microsoft Research group, which designed and programmed an interface that enables Ada to interact with human sentiment. These environments emphasize a shift in responsiveness, where systems are no longer reactive to machines or predetermined functions, but rather to the movements and actions of living entities.

Nicole challenges the label of "flexible" to describe such environments, as it implies a certain compromise or ignorance of the system's potential. Instead, she suggests the term "manipulative" to convey the constant state of change and adaptation that characterizes these spaces. This notion encompasses a range of states and activities, highlighting the dynamic nature of the environment. Negroponte's text also introduces the concept of two distinct types of responsiveness: operational and informational. Operational responsiveness entails visual or auditory responses integrated within the spatial context, reflecting a purpose or intention. On the other hand, informational responsiveness is simulated and necessitates the preparation and acquisition of relevant information to effectively simulate various environments and experiences.

It is worth noting that objects akin to Sabin's Ada project can be considered alive in their own right. These objects possess the ability to exist and derive sustenance from their surroundings, including the sun, human presence, air, humidity, and other environmental factors. This adds another layer of engagement and interdependence within the object or project's entity. Consequently, the design and development of intelligent environments encompass a comprehensive understanding of the intricate relationship between architecture, technology, and human experiences. By exploring the boundaries of personalization and responsiveness, Sabin and Negroponte challenge traditional notions of space, paving the way for novel conceptualizations and applications within the architectural field.

In the chapter titled "This Building Does Not Exist: An Attempt at a Theory of Neural Architecture," the author embarks on an exploration of neural architecture as a burgeoning and potentially transformative sensibility. While neural architecture has found significant application in fields like music and photography, the author posits that its integration into the realm of architecture presents unique challenges due to the
discipline's inherent entanglement with diverse domains throughout the design process. Undeterred by these complexities, the author initiates their argument by elucidating the underlying connections between computer science and neuroscience, seeking to establish a framework for incorporating these principles into architectural practice. By delving into the inner workings of generative adversarial networks (GANs) and
their potential as design tools, the author lays the foundation for a novel approach to architectural creation.

Central to the author's vision is the aspiration to transcend realistic outcomes, instead aiming to engender a sense of both familiarity and defamiliarization. The author astutely notes that training machines using flawed or imperfect images is essential to cultivating their creative potential. Drawing inspiration from esteemed artists such as Mario Klingemann and Sofia Crespo, who have harnessed the power of StyleGAN
in other artistic fields, the author posits the viability of similar techniques in architecture. Thus, with a dedicated team, the author embarks on experiments that involve Baroque plans, seeking to examine how neural style transfer can engender fresh designs that seamlessly amalgamate elements from both Baroque and contemporary architectural aesthetics. The resulting architectural compositions possess a captivating
quality that simultaneously evokes familiarity and innovation.

Furthermore, the author ventures into an exploration of the ontological and epistemological qualities inherent in neural architecture. This entails a profound contemplation of the role architects play in the design process and the knowledge that becomes embedded within the very definition of neural networks. Drawing upon the analogy of a theater, the author deftly elucidates the ontological dependencies that emerge between concrete and abstract objects, as well as the intricate interplay between these objects and the architect's internalized knowledge. These "entities" merit scrutiny by closely examining the nature of project properties. The article expounds upon this concept, identifying two fundamental types: categorical properties, which elucidate the essence and characteristics of an object, and dispositional qualities, which
encapsulate the latent potentialities harbored within an object. It is through the analytical lens of these project properties that the application of AI in architecture primarily focuses on categorical attributes, encompassing elements such as shape, dimensionality, and spatial configurations, which lend themselves remarkably well to neural network analysis. Consequently, neural networks assume the role of abstract
objects, defined by their inherent properties, features, and intricate relationships.

Expanding upon the investigation of ontological qualities within the realm of neural architecture, the article delves into the realm of aesthetics, agency, and authorship. Drawing from a rich tapestry of philosophical, architectural, and artistic perspectives, the article unveils the enthralling aesthetics intrinsic to neural architecture, thus engaging in a thought-provoking discourse surrounding human and artificial ingenuity. Furthermore, the role of architecture in both historical and contemporary contexts is critically scrutinized, as the article contemplates the potential emergence of a posthuman era. The intertwined concepts of free will and agency come under scrutiny, as the article explores their interconnected nature. Human agency, characterized by the ability to make conscious decisions and assert one's will, prompts profound contemplation of moral agency and its associated consequences. In this context, the article references the examination of agency within the realm of action theory, drawing inspiration from the philosophical insights of Hegel and Marx. Through this lens, the relationship between humans and neural networks emerges as a collective endeavor, driven by a distinctive form of agency that transcends the conventional boundaries of individual human agency.