Architectural Strategies Built on Artificial Neural Networks

Independent architectural intelligence that communicates and react to individuals through cognitive interfaces.

 

Studio: FLOW 2017/2018

Project: Sound Shape Space

Author: Karolína Kotnour, Ing. arch.

Supervisor: Miloš Florian, doc. Ing. arch. PhD.

Cooperation: Institute for Research in Science and Art, Robert B. Lisek

Supported from Student Grant Competition: SGS17/143/OHK1/2T/15 FA ČVUT 2017

The project introduces an architecture that represents a constructed expression of mental space and imaginative insight. The fundamental needs and motivations of human are determined by physiological possibilities and biological and psychological conditions and the habitat. The same motivation takes a human to continuously transform the space of the natural environment into the artificial (technological) environment. The architecture is understood as a dynamic organism that is driven by the creation of humans needs caused by the lack of instinctive adaptability to the dynamic environment. Architecture work with physical and virtual materials where virtual material is information in large databases. Machine learning algorithms reveal patterns in natural processes and human behaviours and are incomparably more effective in their meaningful application and immediate recomputing of applied and proven knowledge, a self-encoding architecture for a complex synthesis of dynamic environments.

We propose a solution for problems linked to the creation of dynamic interactive architecture, specifically the problem of continuous adaptation of artificial agents in complex dynamic environments. This refers to the difficulty of designing artificial agents that can respond dynamically and intelligently to evolving complex situations. Environments are considered dynamic when there are changes in the structure of the environment, in the presence of multiple learning actors, or in the presence of human users.

When we look at architecture as we know it now it is a hardcoded system with very limited ability to respond to its environment or changes of this environment, but we design architecture that has the ability to re-code it’s own system and is able to communicate and activate more complex systems and actually continually learn from the environment. 

In this case study, we focus on the continuous adaptation of artificial agents in complex dynamic environments, such as soundscapes, in immersive environments and testing their adaptability in a dynamic environment. The environment is created by using augmented reality, virtual reality, or holographic experience of reality, in which a user will be physically immersed in the environment, and will interact with agents. The proposed model will be more powerful than similar, existing models, and will find wide-ranging applications in Architecture. The ability to continuously learn and adapt from limited experience in a dynamic environment will be an important milestone on the path towards building a General Intelligent Architecture.