Course Syllabus

Jump to Today

GSAPP Link | William Martin

Course Documents

Context

In 2022, OpenAI's ChatGPT and DALL-E highlighted techniques called "generative artificial intelligence," which have since captivated the world with remarkable creative abilities. Simple text prompts can now appear to write essays, perform deep research, generate striking imagery and video, even author and debug code.

At the same time, artificial intelligence (AI) now pervades the built environment. Governments track public spaces; autonomous vehicles and dog-like robots continuously map environments in real time; tech companies listen to and monitor our homes; and satellite data feed predictive models of urban growth and destruction.

The most influential large language models (LLMs) like GPT-4 and LLaMa 2 are currently built by the wealthiest tech companies, raising questions about their futures and incentives. Text-to-image models like DALL-E and Midjourney have quickly revealed the biases pervading today's internet society. But there are also open AI initiatives starting to challenge this hegemony, but on what data should we train future models to ensure fairness and safety?

In a rare proactive trend, perhaps due to lessons learned from social media, governments are acting to regulate and guide further development of these technologies. The EU have been developing an Artificial Intelligence Act since 2019, and only recently have the US issued an Executive Order for AI Safety, contemporaneously with the Hiroshima Process and the AI Safety Summit in the UK.

Despite such urban surveillance strategies and emerging governance challenges, the same enabling AI technologies present an opportunity to spawn sophisticated design methods. And for new spatial AI methods to germinate and grow into their creative potential, computational designers must research and develop new AI models and agents, those tuned to spatial design problems and the semantics of space.

This Course

"Spatial AI" refers to the emerging, inventive applications of AI to all aspects of spatial reasoning, design, and experience. This survey course will:

  • explore the definitions and affordances of generative and other forms of AI,
  • scrutinize canonical writings from relevant theories regarding space and intelligence,
  • experiment with the rapidly evolving landscape of commercial (e.g. ChatGPT, OpenAI API, Claude V2, etc.) and open source AI (e.g. HuggingFace, stable diffusion, etc.),
  • develop a critical and technical understanding of the technology, and
  • speculate on new spatial AI methods at human, architectural, and urban scales.

It is divided into six two-week modules, each with critical readings and exploratory technical exercises.

 

Module

Spatial Concepts

AI Tools + Technical Topics

Introduction to Spatial AI

What is AI? What is space? Classical models of space.

How does AI work? What are the subfields of AI?

Spatial Reasoning in Text

Rule-based reasoning, descriptive space (e.g. building codes, zoning codes, etc.), spatial semantics

Text generation, natural language processing, large language models (LLMs), transformer architecture, prompt engineering, training, inference, fine-tuning, OpenAI API

Spatial Reasoning with Imagery

Image space, place space, authorship, bias in AI

Computer vision, text-to-image, text-to-video, video-to-text, object + facial recognition, convolutional neural networks, generative adversarial networks, diffusion probabilistic models

Intelligent Experiences in Three+ Dimensions

Virtual space, way finding, social space, semantic models

Audio-to-text, 3D data, point clouds, LiDAR, object generation, object recognition, AR/VR, body tracking, AI in games, text-to-object, scan-to-BIM, navigation

AI Design at Urban and Regional Scales

Connective spaces, geographies, ethics and safety of AI

Satellite imagery, graph generation, predictive modeling, spatial tracking

Speculative Methods

Proposing new spatial AI methods

Design computation, computational models, AI agents and agent chains

 

Course Summary:

Date Details Due