Perkins Eastman Keynote

Presentation Titles:

Interconnected AI Ecologies: Towards Successful Strategies in Developing AI-Driven Design Workflows

Paragraph:

Artificial intelligence technologies are defining an unprecedented paradigm shift, changing the discipline of architecture as we know it. Drawing upon the keynote speaker’s expertise in cutting-edge research at the intersection of AI and architecture design and her developed research projects, "Creative AI" and "Performative AI", this session addresses the potential of integrating AI into new design processes. It offers insights, through case studies, into methods of establishing successful frameworks for AI-driven design processes. Derived from our multiple workshops and publications, we share knowledge on developing AI-driven systems and workflows that involve identifying appropriate AI models (i.e., generative deep learning models including CycleGAN, StyleGAN, Pix2Pix, and text-to-image models: DALL-E, Midjourney and Stable Diffusion) for specific design tasks, establishing relationships between these models, and defining connection strategies. From a procedural standpoint, our focus lies in recognizing the correlation between specific types of AI networks and their potential to tap into creative outcomes in a deliberate, heuristic process. Assessing an expanded design space, enabled by interconnected AI models, can transform our design exploration activity, enabling access to a search space that would otherwise be beyond the designers’ reach. The objective is to emphasize strategies to augment designers’ creativity and automate specific design tasks in a hybrid system where interactions among the AI nodes can lead to innovative design solutions with new aesthetics and higher performance. Attendees will acquire actionable knowledge, enabling them to effectively leverage AI tools and foster innovation within their design process.

The session will demonstrate the integration and connection of multiple AI models to address various architectural design tasks, facilitating both creative exploration and the development of performance-oriented AI-driven design systems. Case studies from multiple workshops and publications will illustrate the different strategies developed to formulate successful AI-driven systems with systematic encoding of design intentionality as a continuous series of actions within the interconnected workflow. Attendees will gain a nuanced understanding of methods of dataset curation and connection strategies for AI models, as well as methods of evaluating these workflow strategies, including the linkography method. Collectively, participants will cultivate a discerning sensibility for these processes by leveraging the capabilities of AI tools and employing effective methods to evaluate their developed workflows’ outcomes.