Towards the Matrix...and out the other side

Walking out of the cinema, in 1999, after seeing The Matrix for the first time, was my first experience of virtual reality.  That's right, not watching the film but walking outside.  The city at night: street lights, the uneven pavement, the looming London plane trees across the road, they all quivered with a kind of hyper-reality I didn’t trust.  I had just spent over 2 hours moving in and out of a world that looked like mine but was actually built by computers.  Those people passing me by, were they real? Heads down, hoods up, not making eye contact. Could they see me? See what I saw? Was any of it real or just a projection my mind was locked into?

Years later I found myself musing about the use of AI in architecture and the built environment and where it might take us.  This time I was on my bike, passing a building I’d done the technical drawings for.  It was daytime and a different city but there were still trees and traffic and bumpy pavements.  I wasn’t questioning my reality, but I was wondering where things might be heading. Twenty-six years on and the fundamentals of the cityscape hadn’t changed. Where was the experimentation? The never before imagined forms and spaces?

I looked down at my forearm and at my hand on the handlebar.  Maybe AI in building design will have to be different to music or film or visual arts because we deal with the physical world.  Humans have bodies, I thought, and that’s why we need shelter.  In fact, we are on the very line between the real and the virtual because our minds can inhabit virtual space while our bodies are rooted in the physical.  So, even if the city was to consist of blank boxes onto which we each projected our fantasy dreamscape, we would still need the boxes – wouldn’t we? Which was when I realised what had happened in The Matrix. 

Humans had stepped off that line and wholly into the virtual world, or they had been forced to ( I don’t really want to get into the minutiae of the plot although I accept there’s a lot of fun discussions to be had). 

As far as the individual inhabitants of the Matrix were concerned, their bodies had been unknowingly abandoned to exist in slime filled pods: physical comfort no-longer a requirement, sustenance provided, environment controlled. 

Is this the only way to experience the full potential of artificially imagined spaces?

I would like to be properly excited about using AI in architecture and construction.  I will confess that my understanding of what is currently out there is rudimentary at best and I’m trying to master an overwhelming urge just to look the other way. I feel both fascinated and a little bit dirty when I watch AI videos in my Facebook feed and it took me an embarrassingly long time to spot the fakery in the plethora of over-tall (conveniently portrait orientated) interiors also on my social feeds.

Sadly, the virtual NY of the Matrix looked boringly like everything we already know. The hero, Neo is told that they tried running with something more attractive and enjoyable but the human brains simply wouldn’t buy into it.  Given that AI generates its solutions based on bucket loads of existing data sets perhaps I am foolish to think it would propose radically new building forms or cityscapes.  I tried an experiment.  I typed “garden room extension on 1930s semi-detached 2 story house with green roof and large bifold doors” into www.mnml.ai after a google search for architecture concept design ai.  This was the first result:

I’ll admit, I was taken aback by the apparent realism of this from my very rudimentary description.  OK, so it clearly doesn’t know what semi-detached is and that’s not the kind of ‘green roof’ I meant but boy, I get the seduction of the process.  Five random variations later and all my credits used up, I managed a photo version of something I might conceivably show a client:

So, it’s an amazing visualization tool but there’s another parallel with the Matrix’s computer-generated world, which is that it relies on the human brain to over-lay a sense of reality. 

Morpheus tells Neo that the programme follows the same basic rules as the real world, like gravity, but like other rules in a computer system these can be bent or broken.  Most people in the virtual world still believe they cannot perform super-natural jumps between buildings for example. They impose assumptions based on experience.  Now look at the lower roof on the last image, there’s something odd about it.  Perhaps you could build a roof like that, but I don’t understand how it ties in with the geometry of the spaces below.  I’m assuming the bifold doors fill in that opening, if they were all closed but that also means its an extraordinary wide span for a domestic building – is there room for the depth of steel required above it? and what is actually going on with the walls in the room inside? 

This ‘believable but not entirely accurate’ approach is risky for building designers like me, and for the people we work for.  As with other creative processes we use an iterative approach.  We work up and down through scales to refine the solution. Those paying us for our time would love it if we reached our best and most workable solution immediately, I know.  It’s just that there are so many variables and the best concept idea are mouldable in a way that these visualizations do not lend themselves to being.

Thirteen years ago I set up on my own to work on domestic projects just as my profession was embracing BIM – Building Information Modelling.  I was resolved to learn the next generation of CAD software for communicating my design in 3D in a way that could be fully coordinated with engineers and other consultants.  Sadly this proved too cumbersome for the size of my projects and (happily) I was too busy with new work to do much about it.  But it was also the requirement to commit to so many details right from the start, just like the AI images, which concerned me. Perhaps with practice I would have learned to navigate this but I felt the digital tool was already guiding my hand too much.   

My current ‘go to’ for a first proposal for a “garden room extension on 1930s semi-detached 2 story house with green roof and large bifold doors” is still Sketchup.   

The AI image might win over the client, but if I can’t build it am I doing my job?  It might even make my job harder.  Recently I’ve been fielding images from clients of miraculous storage systems: Nania sized closets under standard flights of stairs, shelving over washing machines which could only have been sawn in half, impossibly tall kitchen cupboards.  It takes work to show why these apparently achievable ideas are either physically impossible or if not impossible, then way out of their budget.

When I completed my Postgraduate Architecture Diploma in 2009, I wrote a thesis on ‘Craft Thinking and Digital Making’.  I argued that there was a place for using digital tools to spark imagination in the same way a particular piece of wood or stone might stir a sculptor.  I suggested that a craftsman or woman employs an instinctive part of their brain which draws on their knowledge and experience and guides them to an original creation and that they could do the same with digital tools without being overwhelmed by them. 

What block of marble equivalent can I throw into my AI tool to generate a completely new concept for a house, a library, a city? How would the iterative process look? 

1.        Take this abstract inspiration and turn it into a form

2.        Simplify it, extend it, push and pull it

3.        Scale it so that I can walk around in it

4.        Smooth it, roughen it, open this part, close that

5.        Apply the laws of physics to the form

6.        Apply suitable materials and building codes

Oh maybe not building codes, not straight away, give me a moment to be truly blown away by the possibilities because if we redefine those boundaries then we will need to redefine our regulations too. 

There are architecture studio heads and teachers who have been advocating this kind of craft thinking for decades so has it infiltrated digital design processes or are we heading for The Matrix mediocre?  Will we find ways to invent truly futuristic forms to live in or keep it bland and only build the exotic experiences to inhabit in our head?  I would like to caution everyone on the deceptive seduction of instant design solutions.  As a speaker said at a recent AI Symposium I attended, the novelty of creating an image from words wears off in about seven minutes. We are at the ‘I can’t believe it’ stage of AI but, just like the Matrix, as mind blowing as it is, when it comes to our built environment we have to demand more.