Aesthetic Judgment Services: or how to make architecture great again with artificial intelligence

This is part of a series focused on technologies and systems.




You create a learning algorithm for recognizing architecture. You feed it millions of pictures of buildings. You have users all over the world answer simple questions about buildings from photographs.

Let's say you have the following categories;

Skyscrapers
Civic Buildings, (museums, court houses, etc.)
Small and Medium Churches
Cathedrals and Large Churches
Large Malls
Box Retail Stores
Strip Malls
Houses
Mansions
Gas Stations
Automotive Facilities, (car washes, auto parts stores, mechanics, car dealers).
Warehouses and manufacturing facilities
Manufacturing Plants
Self Storage Facilities
Refineries
Ports

First every building gets put into a category. The AI learns how do do this as users are presented with pictures of buildings and asked to categorize.

Then two pictures of each kind of building are presented to users who are asked: "which one is better looking? The one on the left or right?"

This trains the AI to have a sense of aesthetic "taste."

You repeat the second step over and over again so that you can teach the AI to have good taste in architecture. Moreover, the AI's good taste is the same as the public's good taste, so most people will like the architecture it approves of.

Now you have a computer program that recognizes good taste for each category of buildings. It knows what a good looking warehouse looks like, a good looking mall, house, skyscraper, etc.

You approach a local government and get them to write into their building code;

"For a building to be approved it must be uploaded to the website run by our contractor, Aesthetic Judgment Services, (AJS), where the AI will decide if it meets aesthetic qualifications for the 1 % most beautiful buildings."

Or if that is too strict then maybe the 10% most beautiful , or the 20 % most beautiful.

An architect can upload his building renderings to the website and receive approval or denial in minutes by email. The AI encodes the aesthetic standards of the community.

The AI can even give feedback on why the design was denied, thusly;

"Too boring."

"More ornamentation."

"Too cookie cutter."

"Are you seriously using Revit windows?"

"Too cheep."

"Brutalism has less than a .001 % pass rate."


Don't think a manufacturing plant can't look good?

What about a gas station?

What about self storage? Do you think that has to be ugly?

What about the humble strip mall?

Churches of course can be amazing too. Even when they are modern.

The point is that you actually can encode aesthetics into law. You just need the proper machine.

Oh, and modern architecture does have its place for things like gas stations and strip malls.

Most of these buildings are still fairly cheep to build too. The storage facility above uses polycarbonate (plastic) for god's sake.

Now some may say, "oh but a king would do a better job and judging things." Oh yes, he might. But until then. . .

There is also the issue of corruption. An AI cannot be bribed, and it's results can be transparent and examined by looking at its code. Humans can have poor taste, weird taste, no taste, and monetarily flexible taste. Machines are neutral.

Moreover, you can have different AIs for different cities programmed with different ideas of "good taste." For example;

AJS: Rivendell Edition

Or AJS: Japanese Edition

You can customize AJS services to your local values and culture.

Think about it.