Michael Bergin

Why this site exists and how it is put together.

Why you still can't use AI to design a better home — and why this is the biggest opportunity in applied research.

We build hundreds of thousands of homes a year that no one is proud of. The cost of care has always been the cost of change — and adaptive homes finally make change cheap.

A six-level framework for AI in the practice of design — from inert CAD tools to guaranteed code compliance — in the spirit of the levels of driving automation.

A new class of deep learning methods can translate automatically between representations — a sketch into code, a mockup into a working site. For designers of the built environment, that changes what is possible.

The core thesis paper — a taxonomy for a generative design system that lets architects, engineers, planners, and clients negotiate the design of a house through Search Constructors, Simulations, and Goal Sets.

A field survey of multi-objective optimization arriving in architecture — Galapagos, fitness functions, and the high-profile projects (SOM, OMA, ARUP) quietly using it. Written 37 days before my thesis defense.

A speculative proposition — what if generative design made a custom house so cheap to produce that it could be given away as a public service, and rated like a restaurant?

From Wilde's defense of lying to Frankfurt's account of bullshit — on art, nature, simulation, and how much of architecture is confident talk about things too vast to fully grasp.

A piece of speculative advertising copy from the thesis — generative design sold to the public as a consumer product, circa the near future of 2012.

Engineering has long applied the demands of reality to fantasy — Gaudí's catenaries, an automobile's fluid dynamics. What happens when it can do the same for a house?

Computational design is, by necessity, complex and contradictory. The decorated shed — Wal-Mart most of all — is what happens when we refuse to say so.

There are more unemployed architects than there are houses built by the largest homebuilders. The talent to customize exists; the system to deploy it does not.

Le Corbusier stated the house as a problem to be decomposed and mass-produced. A century later, the conception has been forgotten and the developer owns the result.