There was a curious development in the 19th century which I call as the episode of Crinolines. It started around 1845 and lasted for almost 50 years before thankfully dwindling and dying out. Crinoline was a bulbous dress worn by women of those times. To create Crinolines, a framework of circular hoops was first created from the waist down. These hoops were made of Bamboo strips, Cane, or even Whale bones. Then, various layers of clothes were draped and stitched on top of these hoops, with intricate embroidery and other designs. The crinolines billowed out from the waist of the women, often to absurd extents, sometimes reaching 6 feet or more. To say they were cringy is an understatement. However, during that period, they were considered high fashion and coveted by the society ladies of the time.
I want to connect this curious episode of the crinolines with the trend that is happening in AI, especially AI in architecture design.
I hope I have piqued your interest and you are keen to know about this episode of Crinolines; and the lessons to be learned for architecture design. That fashion trend was aided by an important invention that came just before the episode of crinolines. That is the invention of the home sewing machine. I call it the laptop of the 19th century. And why so? Just hear me further.
Let me give you some context. This fashion trend was happening in the cold countries: Canada, US and the Great Britain. Cold is a huge influencer in clothing. People in those countries cannot be casual about their clothing. They can easily die if they were not clothed appropriately. So they often spent a lot of time stitching clothes. A task that was quite hard and time consuming. Only the rich could afford to employ tailors and seamstresses. But not the general laity.
Hence; when the home sewing machine came into the picture, suddenly a lot could be done in a short time. This excited the imagination of simple people. Crinolines were spurred on by the fact that people could try all kinds of stitching. Logic and careful reasoning were left behind. They dived deeply into possibilities without much reasoning. Not because the possibility was a necessity.
Crinolines were a disaster in the making. Thousands of women died eventually from it. The billowing dresses used to catch fire at one end, which the lady wearing the dress did not notice. Others could not help her either; as they themselves were wearing crinolines. Ladies would get stuck in doorways, passages and staircases. Climbing staircases was very tricky and many fell easily.
Some ladies were lifted up by the wind and smashed across the street or trampled over by horse carriages. In short; society saw all kinds of unimaginable ways of dying and getting seriously injured. It took society almost 50 years to finally get out of the clutches of this crinoline fashion.
Why would I now use this as a way to connect to the trend of use of AI in architecture design. Some may think I am connecting it to the fashion of generative geometry. After all; fabricating clothes on a framework of support hoops is surely an example of generative geometry. Even though there are some lessons there too; that is a separate point actually — and if I stated just that connection of generative geometry; it could argued that I am making a strawman argument here.
I want to point out something more insidious; which is quite pervasive in humanity right across history. As humans, we are always captivated by the experiential world. Empirical knowledge and the inductive reasoning that we humans do can be quite misleading. The reason is simple: Inductive logic is only probabilistic. It can sweep over individual differences and lead to hasty stereotypes. Especially in the hands of influential but untrained people.
The visual experience especially is a demanding mistress we often bow to. Of all the senses we humans use to inductively detect patterns; it is found that the visual apparatus in our body works the fastest. Now when a majority of society got a tool which can lead to multiple visceral and visual experiences, the impact became much greater. This heady combination of a new laptop in our houses; combined with the need to excite our visual imagery is dangerous. People who are not qualified or cannot put a pause in their excitement, and can’t become pensive can easily jump into a design conclusion.
And that’s the connection with architecture. Our field is full of unqualified and insensitive; so-called “designers” who try to win brownie points by doing visual jugglery.
There are many out there who can wield popular 3D software; who believe they can create architecture using that. Today we have real laptops in our houses — a general purpose tool that can be reshaped just-in-time to suit quite a lot of different purposes. These laptops are extremely powerful and they are getting even more so every day.
The potential that such laptops releases has actually created a lot of wide-ranging problems, but let me not widen the scope of this article and instead stick just to architecture and design. This is where I now bring in cringy AI. I nowadays see a flurry of people with no deep knowledge of AI going around proudly claiming that they could make visually scintillating designs using AI. These people have absolutely no idea of the history of AI – why it came into society, what were the theoreticians talking about, what are the dangers that AI can lead to and many other fine points.
They are like those simpletons of the 19th century with their heads bowed over a home sewing machine, fascinated by the way the stitches are formed with swirling ideas about bulbous crinolines. I get really worked up when people confuse machine learning with the whole of AI. It is like mistaking the simple guesses that a growing infant may make when learning the mother tongue; for the entire intelligence of that child. Machine Learning is one tiny, fascinating part of AI and surely cannot be the whole of AI.
There are two broad approaches to get into Artificial Intelligence: One is the symbolic approach and the other is the sub-symbolic approach. The symbolic approach is based on high-level symbolic (human-readable) representations of problems, logic, and search. It uses tools such as logic programming, production rules, semantic nets and frames, and it developed applications such as knowledge-based systems (in particular, expert systems), symbolic mathematics, automated theorem provers, ontologies, the semantic web, and so on.
On the other hand, the sub-symbolic approach is based on low-level neural networks that are inspired by the structure of the brain. Or to clarify; it was spurred on by a popular understanding of how the brain is structured. Some believed that such a structure influences its working and production of thought. So they borrowed the same and got Machine Learning.
This approach is also known as connectionism or neural networks. It involves learning from examples and it is quite telling that when it was first used in applications such as image recognition and speech recognition. Topics that were to do with our empirical senses. Now it is used not just in recognition but in generation also. If you used ChatGPT and other such tools, some of you may notice that often they can smugly tell you a lie — machine learning often leads the horse to a general-purpose waterfront of possible ideas. But not the exact water to drink from. I have seen ChatGPT and Bard happily spew out bad advice just the same way a Whatsapp University uncle may do so. At least with verbal advice, there are some ways and means to detect such falsehoods. But when it comes to visual imagery; it is really tough to see whether the generated design was based on falsehoods or not.
If you scratch the surface, machine learning is highly reductionist, highly inductive and often loses the meaning of the whole with its fascination for the parts. I am an avid critic of this sub-symbolic approach in AI. Not just me; but even important people who used to work in this area are now retracting from their earlier stances. For example; Geoffrey Hinton, sometimes considered as the father of machine learning, is now going away from his past work in that field. The simple truth is that machine learning is currently quite hyped up, often to serve devious commercial agendas. And to hell with accidents waiting to happen.
When I got involved in Prolog and I was shaping my own design software way back; starting from 1987 onward, I was fortunate not to be caught up in this hubris and hype about AI. In fact, AI was simply a different approach to computer programming and almost everyone agreed those days that the “Intelligence” in AI had absolutely nothing to do with human intelligence. There was a lot of nice research into linguistics, use of symbols and so on. Today I am quite convinced that sub-symbolic approach of AI without first working out the symbolic nature of representation of knowledge would simply lead to disasters. Of magnitudes and kinds that humanity may have never seen before.
We forget that 9/11 happened just a few years back –it marks the advent of a new juxtaposition of politics, religion, unfairness, envy and architecture — a combination that society had not seen much in such a virulent form.
AI of this flawed nature is surely going to throw up even more strange juxtapositions. It is now given in the hands of many thoughtless people; even smug ignorant architects too. Just because the simple AI tool user is not really an AI researcher, they are like the simpletons who were fascinated with the home sewing machines. Both are fascinated and deeply captivated by the imagery they can lead to without realizing the true, deeper merits of such home laptops.
Let me conclude by saying I am grateful for the invention of the simple wheel: The circle by itself is quite a boring, abstract figure; with no adornments. In fact, if one tried to place decorations on the circumference of a wheel, it possibly would not have worked as a device for locomotion. So the circumference of the wheel had to remain simple, and it still is largely remains simple. With the visual imagery of the wheel thus supressed, people could delve into other properties of the wheel … and the wheel went on to uplift humanity. Unlike the crinoline, which sadly lifted up the entire human being and dashed her across the street and killed her.
Some useful links:
What Was Up With Those Giant Victorian Skirts?
Featured Image: Building form inspired by crinoline–Created with Bing Image Creator.
