DANIEL BOLOJAN
PROJECT
AgentBased CreativeAI’s
AgentBased Creative Ai research envisions the potential disruptive change in paradigm driven by the adaptation of creative AI methods in architecture and how these can augment designer’s native ability to solve against design problems. Can AI truly support designers creativity and discovery?
Design space explorers - while designing, in an abstract way, one will start from an empty canvas, an empty design space, or solution space that one will try to gradually narrow down to something that they can work with as a starting sketch. Each decision taken will open new design possibilities while closing other possibilities. However, most of the time, the designer is unaware of all the design possibilities at a given point in the design process or when choosing one possibility, if that is the most relevant outcome (Local minima vs Global minima). Design intent is used here to describe architects/designer desire, sensibilities or design criteria. It can refer to aesthetics, form, structure, program, tectonic articulation, affects or any other aspects of architecture.
Once the design intents are defined one might ask, which are all the design variations for a particular design space. Design space is defined here as the multidimensional combination and interaction of design input variables and design process parameters that represent the design intent. We are familiar with the expression "Hunting a needle in a Haystack". I believe the design process is better described as "Hunting a needle in a flock of birds" that constantly changes.
The design process is better understood as an ever expanding and ever contracting space of possibilities and possible outcomes. At each given point, in the design process a manifold of design intent variations will be generated. A Neural Network is used here to reduce the high dimensionality of data (multiple design intents defined by parameters) whilst also will retain the high-dimensionality, non-linear associations between multiple design intents. The resulting design space can suggest an overview of possibilities within the given parameter space, associations and disassociation of design intents. The network creates a non-linear interpolation between multiple design intent inputs and a location rule by placing similar design intents closer in the design space while placing further apart dissimilar design intents.
The idea isn't to start from an empty canvas, but rather to start with a multitude of options and through a process of filtering and selection, pick one and zoom in from there. This approach represents a shift from the idea of a design as output, to the idea of design as multiple outputs, while the designer is engaged with a process of selection.
The role of Creativity in this design process is in defining the constraints that generate a range of possible solutions for a design problem, define and develop strategies and effective methods of filtering and evaluating possible solutions. It's no more a question of defining the ideal design, but rather of reviewing the entire design space of possible sample outcomes, selecting and enhancing the best solution.
– Daniel Bolojan
PROJECT VIDEOS
AgentBased CreativeAI - Structurally Encoded Agents
AgentBased CreativeAI - Qualifying Criteria’s
AgentBased CreativeAI - Design Space Explorer
AgentBased CreativeAI - Representation Learning
PROJECT
DeepHimmelb(l)au
While developments in Ai mean computers can be trained on certain creativity criteria, the degree to which Ai can develop its own sense of creativity is still something to inquire about. Can Ai be taught without guidance how to create? Can Ai be taught how to interpret things? Can Ai be taught how to reinterpret representations from one domain to another, similar to how architects are inspired by concepts outside their architectural domain? Teaching computers to be creative is inherently different from how people create, but we do not yet know much about our own creative methodology.
Our perceptions and our conscious visual representations of reality are not a direct mapping of the real world. Humans interpret reality through reconstructions and interpretations based on past experiences. Our past experiences act as a frame / filter on our way of interpreting, understanding and perceiving the real world. Our training as architects operates as a filter / frame in the way we perceive the world, the way we interpret it and the way we draw inspiration from it.
One very common practice in design and architecture is that a designer learns, consciously or unconsciously semantic representation of one domain, reinterprets that representation through a particular filter e.g. architectural style, architectural culture etc, and translates it to a different domain.
While humans unconsciously are capable of recognizing and disentangling various semantic features of what they perceive, neural networks are capable of having similar behavior after learning from a large enough set of samples. Some Networks learn automatically to separate/disentangle various semantic features of a dataset and afterwards enable specific features to be separated and managed on a particular level. In addition, machines exposed to large sample sets can discover perceptual deficiencies in human recognition capabilities. Can this innate capacity augment the creativity and interpretation of the designer?
What is DeepHimmelblau?
DeepHimmelb(l)au is the result of the cumulative research effort undertaken by Coop Himmelb(l)au which operates at the intersection between architecture, practice and Ai/deep learning.
DeepHimmeb(l)au is an experimental research project led by Design Principal Wolf D. Prix, Design Partner Karolin Schmidbaur and Chbl’s Computational Design Specialist Daniel Bolojan, which explores the potential of teaching machines to interpret, perceive, to be creative, propose new designs of buildings, augment design workflows and augment architect’s / designer’s creativity. DeepHimmelb(l)au is currently the most advanced research dealing with the design potential of AI/deep learning undertaken by any architectural office.
What is DeepHimmelblau’s main aim?
Marshall McLuhan had a very interesting comment about the relationship between the creator / designer and his operating medium / tools -”First we shape our tools, thereafter they shape us”. Similarly the research enquires about the future impact of Ai on the role of architects/designers and the relationship between new technologies / tools and designers. What role should Ai play in the design process? Should the role of Ai be to replace architects/designers? Or should it have a design assistant role to interact with designers/architects to augment design workflows and creativity?
Team:
DeepHimmeb(l)au is an experimental research project led by Design Principal Wolf D. Prix, Design Partner Karolin Schmidbaur and Chbl’s Computational Design Specialist Daniel Bolojan
PROJECT
Gaudi+NeuralNetworks
BIOGRAPHY
Daniel Bolojan is an Assistant Professor focusing on the application of computational design and one of the leading voices in the implementation of deep learning strategies in architecture and architectural design process. Over the years, he has taught several design studios and seminars at the Institute of Structure and Design-University of Innsbruck, Florida International University Miami and conducted numerous international workshops and conference workshops, dealing with the application of complex systems and Neural Networks in architectural design.
He is currently a Ph.D. candidate at the University of Applied Arts, Institute of Architecture, Vienna – Austria. Daniel received his B.Arch. and Master's Degree in Architecture from the University of Applied Arts, Institute of Architecture, Vienna – Austria, where he studied under the late architect Zaha Hadid and Patrik Schumacher at the Zaha Hadid Vienna Studio. He later joined the research project “Agent-Based Parametric Semiology” (Research Grant Funding- PEEK - FWF. Der Wissenshaftsfonds) as Research Fellow under the supervision of P.I. - Patrik Schumacher. The research explores agent-based systems as agent-based life process simulations (architectural crowds) in order to operationalize the semantic layer within the design process, where the semiological code is defined in terms of the agent’s behavioral rules when interacting with a variety of spatial features.
In 2013, he founded his own research studio Nonstandardstudio. Over the years, through Nonstandardstudio's work, Daniel’s design research developed at the intersection between generative design, computation, multi-agent systems, neural networks, deep learning, and machine learning. The studio focuses on generative design strategies and algorithmic techniques that target the creation of highly complex autopoietic systems that could offer new opportunities for the architectural organization, articulation, and signification. These strategies emerge from growth processes, rule-based, multi-agent systems and bottom-up driven design.
Upon graduation, Daniel joined the internationally renowned architecture office CoopHimmelblau, Vienna – Austria, as Computational Designer. There he had the opportunity to practice on numerous internationally renowned projects and competitions. Shortly after joining CoopHimmelblau, Daniel held the position of Junior Associate, Computational Design Specialist & Founder and Head of Chbl|Code. As Head of Chbl|Code, he held the leading role of developing custom computational design tools (e.g. standalone apps, plugins, and add-ons), computational design strategies, virtual and augmented reality applications, machine learning, and neural networks applications, as well as robotic fabrication processes. He is responsible for the office’s current drive to develop deep learning strategies aimed at the augmentation of the designer’s native abilities through the development of the DeepHimmelblau Neural Network.
