3D Neural Synthesis: A Workshop on Architectural Intelligence 

‍

Dates: Starts October 1st 

Registration: $199

Instructors: George Guida – ArchiTAG; Carlos Navarro – OfficeUntitled LA; Daniel Escobar – Online Lab of Architecture (OLA)

Tooling: 2D and 3D AI, Rhino, Unity, Google Colab

Max Number of Participants: 15

3D Generative AI is a comprehensive 4 week course on the most recent advancements in artificial intelligence in architectural and creative practices.

This course centers on the latest machine learning innovations, exploring the cyclical tension between conceptual, digital, and material through the lens of multimodal 2D and 3D form generation. Students will gain the skills and experience in navigating between text, images, 3D models, and immersive environments to create customized 3D artefacts and buildings. We will cover models such as Midjourney, SDXL, ControlNet, Runway’s Gen1, and most importantly 3D Neural Radiance Fields (NeRFs).

Text-to-image models are paving the way for text-to-3D models, thanks to the development of 3D neural networks such as NeRFs. This marks a shift from the necessity of incumbent 3D datasets to diffusion-based 3D models. Through the use of custom scripts and workflows, NeRFs will be used to gain greater control, agency, and authorship within machine-learning aided design processes. The commonly associated anthropomorphic view of AI as competitive agent will be reframed as a collaborative tool that can open up design possibilities and produce new forms of design and digital exploration. Outputs will be shaped with varied levels of creative control into 3D architectural forms using NeRFs and experiencing these through an agent-based Unity workflow. Students will use a hands-on approach to hybridize and create new 3D forms and immersive digital urban experiences.

‍

Learning Outcomes:

• Gain insights into current generative AI tools relevant to AEC industries;
• Generate and customize text-to-3D models using Neural Radiance Fields and diffusion models;
• Contextualize 3D models in selected urban sites using the Unity gaming engine and Google Maps API;
• Understand the potentials of ML agent systems and how these can automate and inform 3D form generation.

‍

Session 01: Sunday 1st October - 1pm EST - 5pm EST

Lecture on Generative AI in AEC

Workshop on 2D image-to-image generation from 3D geometry + Intro to Google Colab 

Images Review + Office Hours

‍

Session 02: Sunday 8th October - 1pm EST - 5pm EST

Technical Lecture on 3D Generative AI 

Workshop on Unity + Google Maps API + NeRF dataset creation and preview

3D Concept Review + Office Hours

‍

Session 03: Sunday 15th October - 1pm EST - 5pm EST

Lecture on 3D Generative AI + World Building

Workshop on 3D Neural Radiance Fields + Unity world building with ML agents

3D Concept Review + Office Hours

‍

Session 04: Sunday 22nd October - 1pm EST - 5pm EST

Guest Lecture on 3D Generative AI and Design

Final Student Presentations

Final Office Hours

‍

Level: Beginner

Format: Virtual (Recordings will be made available during the duration of the workshop)

Required Materials: Each student will be required to have a laptop or PC with the following software: Rhinoceros 7.0 with Grasshopper, Unity, and a google email to access Google Colab. Plugins, workshop files, and colab scripts will be provided to each student at the start of each day. 

Max Number of Participants: 15

Instructors: 

George Guida is a research associate at the Harvard Laboratory for Design Technologies, and co-founder of the design practice ArchiTAG. His primary research focuses on generative AI, human-computer interaction (HCI), and extended reality (XR). He has conducted numerous conference workshops and lectured internationally on topics relating to creative applications of machine learning in design practices. He has completed his graduate studies at the Harvard Graduate School of Design, and worked as an ARB RIBA architect at Foster + Partners and most recently at the MIT Media Lab.

Daniel Escobar is an architecture designer and 3D generative AI researcher, passionate about pushing the boundaries of creative workflows and enhancing the creative process. His primary focus lies in exploring new applications for 3D generative AI technology to revolutionize the field of architecture and beyond.  

Carlos Navarro is an architectural designer at OFFICEUNTITLED, with a Master of Design Research from SCI-Arc and a Diploma of Architect from the Pontifical Catholic University of Peru. His expertise lies in integrating generative AI, CG animation, and simulation technologies to create innovative architectural typologies and styles reflective of the interplay between digital and built environments. He has taught computational design at DesignMorphine and Universidad de Los Andes, Bogota, and run workshops at conferences such as CAADRIA and DigitalFUTURES.

‍‍‍