Department:

Synthesis of Architectural Design

Course Leader:

conf.dr.mat. Anca Vitcu

Teaching language:

Romanian

Learning outcomes:

The course proposes an introduction to the field of Artificial Intelligence (AI) as an innovative and effective partner in design and analysis, offering a new form of assistance in finding the balance among creativity, constraints and complexity.
We will focus on Machine Learning (ML) and Deep Learning (DL) techniques - especially on generative models, such as generative adversarial networks (GANs) and diffusion models - which are increasingly used in generating architectural images, in analyzing performance and exploring new opportunities in design.
Instead of treating AI as a "black box" tool, we will emphasize conceptual and critical understanding.
In the context of architectural design and urban planning, "deep learning" (Deep Learning) can be applied to analyze and understand architectural data, generate new forms, optimize building performance and assist in various aspects of the architectural process. Rather than replacing creativity, we will show how AI can amplify it by rapidly testing a large number of solutions, by analyzing and optimizing them in relation to space, light, structure, or movement.
Many students already interact with AI algorithms through design technologies, often without realizing what makes them work. At the heart of these technologies is Deep Learning, a way in which computers learn from vast numbers of images, drawings, and texts to come up with possible solutions. Models like GANs (Generative Adversarial Networks) learn by studying patterns: proportions, styles, materials, spatial compositions. They don’t “invent” in the human sense, but recombine learned knowledge in surprising ways.
GANs are particularly interesting for architecture because they generate through dialogue and critique. One network proposes an image, while another evaluates it, bringing the result closer to what seems “real” or meaningful. This process is not so different from the culture of the studio: proposing, receiving feedback, refining. Understanding this helps students see AI not as a mysterious tool, but as a system shaped by data, choices, and design intent. The quality of the result depends on what the model has learned—raising important questions about ethics, originality, and cultural representation in architectural design.
Learning the basics of Deep Learning helps architects move from using AI tools to modeling them. Instead of accepting generated images ad litteram, students can critically evaluate them, customize workflows, or even train models on their own design language, local contexts, or environmental goals.
In a future where AI-generated images are everywhere, architects who understand what is happening behind the tools will make a difference – not as image makers, but as architects who can coordinate AI-based technology with responsibility and vision.
We will also discuss issues of ethics, sustainability and responsibility in the use of AI - the emphasis is on responsibility: AI systems must be safe, robust, compliant, and operated transparently, under human control to complement the creative expertise of the architect.
We will understand that AI is integrated through procedures, not used chaotically, we will discuss:
- the limits of algorithms and the risks of using data,
- algorithmic errors and hallucination (results that sound plausible, but are factually wrong, misleading or unrealistic)
- the social impact of automated decisions in the built environment.

Content:

Weeks 1 | AI as a Design Partner
Topics
• What AI is (and isn’t) in architecture
• ML vs. DL vs. GANs vs. diffusion models
• Behind dedicated tools
Week 2 | Data = Design Intent
Topics
• Datasets as precedents
• Bias, authorship, and curation
Weeks 3 | GANs & Architectural Language
Topics
• How GANs work (Generator vs. Discriminator)
• Why GANs are good at “style”
Weeks 4 | Diffusion Models & Control
Topics
• Diffusion vs. GANs
• Image-to-image, inpainting, control nets
Week 5 | Generative Design & Optimization
Topics
• Constraints and fitness criteria
• Evolutionary thinking
Week 6 | AI for Analysis (Not Images)
Topics
• ML for performance and urban patterns
Week 7 | Ethics, Authorship & Practice
Weeks 7 - 14 projects development

Teaching Method:

Lectures based on examples and case studies. Debates.

Assessment:

Project development (individually or in a team)