I interviewed a Design Professional who works at William Rawn Associates, in Boston. He also teaches at the BAS and sometimes at RISD. He completed his master's in architecture and had done a triple major in computer science, economics, and art. He started a consulting firm, with one of his friends and classmates, that focuses on looking at AI in architecture and design.

SUMMARY

The individual is passionate about architecture, with a focus on security architecture. On the technical side, he expresses a strong interest in AI within design, aiming to integrate it into workflows while developing a fundamental understanding of its building blocks. Regarding time allocation in the design process, he would use saved time for tasks like grunt work or exploring more design options iteratively. He expressed concern about the increasing pace of architectural projects and desired more time for thoughtful consideration and reconsideration in the design process. In their free time, he would enjoy activities like music and express a desire to maintain separate hobbies to stay well-rounded. The definition of design, for him, involves synthesizing complex information into a singular entity or artifact effectively. The interviewee believes anyone can be a designer, emphasizing the value of design education. 

Regarding image-generative AI, the interviewee acknowledges its potential but expresses concerns about image saturation and the need for a slower pace in architectural representation. The interviewee has used AI as a tool, particularly for control nets, and views it as a collaborative entity in the design process. He stresses the importance of transparency in AI collaboration and discusses the philosophical and practical aspects of ownership regarding generated images.

If not in the architectural/design field, he would consider finance, tech, or pursuing art. He anticipates a future where a mix of various interests is essential, potentially involving academia.

What are you passionate about?

I think several discrete things. Passion's a good reference, I also like current obsessions. But I think the one thing that unites a lot of them is being on the fringes of my research within architecture. Military bases and survival bunkers have been the longest-running thread with a unifier being security architecture overall.

So, my thesis was focused on that and I had done research trips and I'm still working on a piece on that topic and in that vein. On the other side, on the technical side, I've been very interested in AI within design more broadly. And I think with diffusion models coming in, it's been a good time to bring that passion back fully.

I've been really excited to both figure out new ways of bringing tech into workflows, but at the same time, also just building out a fundamental understanding of its building blocks. So that it can be reorganized, repurposed, and reused and also analyzed and critiqued while it's happening. So I think those are like the two sort of separate chunks that I don't actually have an in-between road between them yet, at least.

If you can save time in the design process, how would you reallocate that time?

Interesting. There's like the Keynesian idea that as soon as you reach a certain level of productivity, you'll be done with something. And then we would all be able to just reap the benefits of our productivity and relax, which is not so true.

And we're still working 40 to 50 hours a week. I am of the camp that whatever time gets saved is just going to be filled in with something new, which is still exciting. I think there's like a separate question about Labor practices on top of that, but assuming that it's time that I'm able to use:

One, I would use it on a lot of the grunt work that I'm not very pleased about in both practice and actually when you're working through projects and design separately, and even academia of just things like setting up line weights or copying the same room tag over and over again.

Labor that does not really require much thought to go into it in the design process. And if I were to gain that time in another way, I think it would either be one actual free time, like I was saying earlier, or wielded back to just try out a bunch of more options and iterative forms of design and in many ways, the one thing I've actually been concerned about in practice has been how fast projects move overall. 

I came to this field of architecture with the explicit interest of it being slow. It's not very fashionable. A project takes six years almost, in some cases, or even longer. And for that reason, it's like a pace that appreciates detail and concentration.

My concern is that everything's shifted towards faster and faster delivery. And with that, we have a limit on the attention we can give, which doesn't collapse as easily. So I would love for that time to be gained in such a way that the actual process doesn't collapse and shorten, but instead, there are just more pauses and spaces to reconsider and to think throughout.

If you weren't putting that time into the project, what would you do with it?

Then a side hobby, one being music, which has been always nice as a fully separate thing. I realize that it's like difficult to actually keep doing architecture.

After a certain point, the well runs dry, and just the idea of not making your hobbies profitable, finding things that are fully discreet from architecture, and becoming a full-rounded human again has been important in that way. So, I think, music and possibly getting back into jujitsu or some physical things again.

Very much lacking when you're in architecture school. What I realized, is you just become a designer fully in every sense.

How do you define design?

I'll come around this in 2 ways. 

I was always interested in design actually 1st from car design. I've always wanted to be a car designer growing up when I was a kid. The thing that drew me to that that I think has still been true is just the unique blend of form and function, especially for car design, there's no way that form can exist without the underlying function, it still has to perform aerodynamically, it still has to work with all the systems that are integrated within it, it still is an object that is very precisely engineered. You think for regular cars, not track cars, pedestrian safety, and crumple zones being a feature while still looking elegant and being marketable so that blend of form and function. When I was younger, I thought it was like art and science and was always interesting.

I then got drawn into design just because, with my background, I was interested in interdisciplinary things, like a blend of disciplines and things that operate in the periphery. And the piece that one professor told me, Alfredo Tierman, actually this,  just paraphrase it, it's just a design of synthesis.

It's not only synthesis, but, I personally think it's just mainly synthesis. It's the act of taking a bunch of discrete things or elements or a complex web of information and synthesizing it into either a singular entity or an artifact, like a design that we produce or being able to communicate it effectively.

So. There are the tools of abstraction that we use, and the methods of representation that we use are all whether it be describing an architecture project that only exists in our head, in school, or in practice, an architecture project that really only exists in our head until it gets built by someone else.

There's an act of synthesizing a very complex thing, for instance, building all the systems and integration that has to go into it and wielding it to become a singular object that can handle that. So with that, I think it's like Design thinking has been a thing I've been much more interested in, and especially with our consulting practice that I've been working with of artifice, taking that kind of method of analysis and method of thinking and applying it to other things, especially things that don't necessarily get a design treatment Today

What makes a person a designer?

One is not being licensed, as a joke. There's like a legal requirement by the AIA to not be able to call myself an architect. But I actually, I've been much more, I don't actually like restrictions on professions, as far as titles, that don't actually have a duty-bound obligation or a risk.

Which architects, doctors, and other professions do. For designers, I think everyone is capable of it. I think there's always been a contrast between design and engineering, at least coming back from STEM fields, like computer science, which I think is worthy of being eroded. My dad, the way he convinced me to keep going with a computer science degree is how creative he told me it is, which was true.

Like you're manifesting something, you're creating something. In that act of creation of a piece of software, there is design choices that have to be done at every step of the way. So, rather than a person discreetly becoming a designer which often has its sort of social gravitas and like coolness to it, since it was invented as a term especially as it got reified by the Bauhaus

I'm open to wielding it, but I don't think it's actually a thing that you have to jump into to become a designer. I think everyone's capable and does design in most of their acts day to day. 

Can anyone be a designer?

I would say that, but I also think that there is value in design education.

Having gone through the process and the ways of teaching design, especially teaching design to people who've l for the quote-unquote, for the first time encountering it, like at the GSE is design discovery there is a value to structuring exercises. To kind of elicit a way of thinking that we all are slowly developing within architecture, over the course of our career, because I think that that's the design doesn't necessarily come naturally all the time, but also it also doesn't have usually have the time and space to thrive in processes.

You're being rushed to do something you might not have time to zoom out and think about it on a design level. I think that a rigorous system of exercises and ways of analyzing the things that you create, which is a product design is still a necessary piece to move in that direction and to exaggerate how often you do it.

What are your thoughts on image-generative A.I. being able to produce beautiful images in a short amount of time?

I think it's complicated when I think. We've seen it actually come in for practice already which, if I were to guess, we would not be one of the earliest firms to do it.

I gave a brief lecture to the firm about it. And so I can pick up around that same time and it's been good for some of our performance health projects as an early design, like instigator of ideas. It's not great when you're trying to find actual forms to be working with. I still think that's just like a product of it being in its early stages of development, and it does not have an actual, like, form and substance underneath the rendering to develop in.

The piece that's within architecture that is unique is that when we produce forms there, there's such a level of specificity and discreteness to the objects we produce. We produce drawings more than we produce renderings. And they're not particularly equipped or beneficial for that, especially when you're trying to inject reality.

But I don't see them. I find the same qualms with us producing renderings also. If we produce renderings before we know exactly how thick a slap has to be, we're gonna run into issues because there's a buy-in of reality and a falseness because this gets closer and closer to reality.

So I don't find them to be fully discreet from that. I think the main concern that I have with them in practice right now, beyond just them being used, to create things that can actually be created, which I like -- I like world-building -- is just the amount of image saturation that we're already dealing with and having more of that.

I think like back to the slowness of architecture. I don't like fastness with architecture and representation. I like how slow things move and have thousands of images, millions of images, be bombarded at you and constantly move representation rather quickly. I don't think as productive as separating yourself from it and working at a slower pace.

So I think we're just going to hit a point where you just have to break away from that and not be starry-eyed by it, which I'm still susceptible to because they're beautiful images. 

Would you use AI as a tool in the future?

I have used it. My favorite use of it is actually, to break from that point, is control nets.

So anything where you can feed it a sort of scaffolding to work off of, either be a depth map or a skeleton drawing or a line drawing and then being able to produce renderings off of that, I think that like locks down some level of specificity, which is what in Artifice we've been very obsessed with is specificity and having some kind of control in the collaboration.

But in that use case, I've been very gung-ho about it. It's like a good way to get material and atmospheric effects. I still think it's very valuable in that, I'm actually fully open to its use both in my practice and in class as long as it's explicit when it's being used.

That's the biggest, just in plagiarism with having a citation for something that you've brought in, declaring that you've collaborated with an entity, which in this case is a text-to-image model is all that's necessary for bounding it and having an understanding of what its drawbacks are.

I genuinely am bullish in that it is going to be impossible to avoid rather quickly, like, in a matter of year. 

When a picture is generated using AI, who does the generated image belong to?

I actually think that the Directorcy is actually a really good way to capture it as an analogy. Because in practice, depending on how a firm is structured there is not necessarily a director, and it's much more designed as a collaborative act.

And even when there is a quote, unquote, 'director,' it takes a committee to design something to some degree, whether it be you, the client, or the 4 people on your team, and even in school, there's that back and forth between an instructor, the students work and the students with each other.

And then also the jury, like, who gives you feedback and anyone who looks at your post on Instagram and comments, all of these acts, if you were obsessively drawing boundaries between them, we would be at a standstill. And I think that saying that though and this might be more of an occult idea around AI that I've been very passionate about is just treating any of these models as an entity.

And, I'm not of the belief that like suddenly we'll reach a point of AGI and then something we can call it a sentient being. I will talk to my car like a sentient being. I think anthropomorphizing and humanizing is actually a useful thing when we're thinking about our connection with things that with tools and technique.

And in that way, I've been much more interested in working through it as a collaboration, rather than as a single directional command. So, you just offer it, 'Hey, go do this.' Instead, you find pieces and moments in a design process and workflow to pass off the torch and then get the torch back between you and the model.

And I think of that as a much larger, like, web of stakeholders that even includes your Rhino modeling software tools that enable being actually considered in a cult sense, like actual entities and persons to some degree that then you collaborate with and work with at every stage so, but in saying that on a technical front, we've found it very productive to splice it and treat it as like this intermediary how we've in Artifice structured it and how we've thought about it.

Theoretically, is it being a mediator for the archive the larger looming archive that exists both your own personal archive, the greater archive, and the collective archive. That has come into both being a helpful thing that can filter through what you are interested in as a companion, which this idea of the AI companion tool -- like a clippy kind of has been pretty pronounced with larger tech companies -- but having something that can sift through millions of images that you've been thinking about and bring like a collective synthesized entity is a very useful thing.

And then you do the active. Selecting just like you would with 15 iterations that you do, and then you feed it back, you do another set of iterations, you check. So that already existing thing of producing a bunch of things and then having a critical gaze to analyze and choose. That back and forth can have some things speeding up that process, which would be a text image AI model, for instance, or various other diffusion models or other visual oriented neural networks including GANs, which I miss and I think have a value for their slowness comparatively.

With that. Ownership comes in the same question of what is the archive is being pulled from the training data in this case, particularly, which I know has been a big piece for artists, including some of my friends who have had their work trained on and be embedded into the model and not getting into the legal arguments for and against it.

Philosophically, I've been less troubled by it in the same way that we constantly pull from precedent and synthesize material and culture just being an act of synthesizing what already exists.. Look in the actual concrete sense, like knowing who you're collaborating with and what precedents are pulling from is essential.

It's how you choose project partners to work with, who you or your teammates are, and know what their references are to some degree. So, having that kind of transparency I think is critical if you're going to be engaging with it on a collaborative level. Or at least a greater sense of transparency that already exists.

For instance, stable diffusion trains on the Leon 5 dataset, or at least former versions of stable diffusion have, and all of the work that is part of their aesthetic dataset, which is the question on their training data of, is this image pleasing to you? Is this image beautiful? And if it scores a certain score higher, like five or above, for instance, then it gets part of that training set.

So for that reason, it has an beyond like all the problems that arise from that, obviously, it also just skews it in the direction of sci-fi fantasy landscapes and certain images of people and really cute animals, and has a very implicit guidance towards certain aesthetics. In this case, what the quote-unquote emergent aesthetic is from having that kind of qualifier.

So, you wouldn't know that when you're collaborating with someone, like, seeing their Pinterest board and knowing that they just are obsessed with things that are the colored blue, you would know that rather explicitly. And I think we can still, like, understand that by just testing things out.

But for that collaboration to work, I think it requires greater transparency to know what kind of model you're working with.

What would you do if you were not in the architectural/design field? 

Good question. I was at an impasse. There are three directions after undergrad, which were: one finance, two tech, and then third, being, an artist, especially specifically in sculpture and installations.

And those three routes still sound pretty nice. But between those, I think it would be, I don't know. I actually really like a variety of things and I actually don't see any one thing in particular, but instead a litany of various interests coming together in a weird hustle, which I think is going to be more and more the case for our generation, which is losing like quote-unquote professions.

Yeah, so we'll see but I think it would be one of those. Some kind of mix of all the various interests and hopefully a foot in academia is a way of housing that.