Get ready for a dive into a world where architectural practice creatively collides with the realms of Generative Design and AI!
Curated content, insights and opportunities to accelerate your career in Architecture, Design, Development and Real Estate to the next level.
Algorithms & Aesthetics: Using Generative Design and AI tools to design Architecture, ft. Arka Works
===
[00:00:00]
Stephen Drew: Everyone put down that Revit model. Go hide somewhere in the office. I'm not going to tell you boss. Don't worry about it. Cause we've got some cool stuff coming up. If you're in industry doing architecture, we're going to talk about AI. We're going to talk about all that cool stuff. Things that you can actually use.
I know, right? Not just a bit of chat GPT. It goes far more than that. All right. Ten seconds. Hello, everyone. Welcome to this livestream special. I'm Stephen. Hopefully you know who I am. But if you don't, stick around, because if you're in [00:01:00] architecture, This is going to be a really interesting one for you. Why? We've all heard about chat GPT. If you're not living in a, under a rock, you probably heard about AI and we're starting to get to grips with it, but how can it be used in the architecture profession, how is it properly used?
How can it be brought in? Now I have some ideas on it, but it's been a long time. Since I've loaded up a Revit model, I used to use MicroStation. That's how much of a dinosaur it is. So it's good to have industry experts that know this sector and know what they're doing. And on that basis, I have the awesome Keir from ArcaWorks, who has just set up this awesome business because he believes in it so much as well.
Keir, welcome to the stage. How are you, sir?
Keir @ Arka Works: I'm good. Thanks for having me. It was a really lovely introduction.
Stephen Drew: I love what you do, but I've got to put the fan, aside. I love embracing the tech. Now, some people in the audience might be [00:02:00] thinking about it. Some people might be sold. Some people might be like, I don't want any more tech. But before we even go into that, Keir, do you want to tell us briefly a little bit about who you are?
Keir @ Arka Works: Yeah, sure. So I'm an architect. I've come out of practice. I was at Morrison Company. I guess that's the most notable I was there for the last 10 years. Last five as a director and decided to set up my own thing this year. And I've gone pretty deep down the AI rabbit hole from around.
November, December last year. And what I'm finding is that there are just more and more things that can be done and interested people and things are moving really quick. It's very exciting.
Stephen Drew: I love it. Yeah, I just think no matter what the business, AI can be really useful, even a little bit, just to do the mundane stuff. However, it's always fascinating, as yourself, you've built these buildings before, you've got the blood, sweat and tears, and now we're going into this new age. I'm pretty excited to talk [00:03:00] about it. I know you've got a presentation as well which is cool and we were going to float around it as well but, okay, what would you like to do? Do you want to just talk about it a little bit further or do you
Keir @ Arka Works: Yeah, sure.
Stephen Drew: show people a bit if you're good?
Keir @ Arka Works: Yeah. Yeah. I think it's interesting that you said, you're using it already. A little bit chatty BT. I'm interested to hear what you're doing. Cause my hunch is that there's a lot of covert use of AI in practice. And one of the things that I think is a challenge is If it's being used in your practice and you don't know about it that could be a thing.
My thing is you should probably know what you're using and why. So how are you using it?
Stephen Drew: I tried to do it for all the mundane stuff or to springboard an idea. So even this title that's a bit inception, even the title of this episode, I didn't use it all, but AI gave me the initial idea. I was like, Ooh, I quite like algorithms and aesthetics. That's cool thing. And it just helped me get past that blank slate.
It helped me get ideas. The other thing I do use it for sometimes is [00:04:00] mundane. Data entry, mundane summaries of certain things. I think it can be useful as well. So sometimes I use it for some YouTube videos I've done to bring out the key points and it will watch that video, bring out the key points and I'll go, Oh, you missed the point there.
But yeah, that's good. That's good. And therefore it just saves me time on doing stuff, which would basically I'd need to do, but you're bit bored doing it. Is that the same for architecture then,
Keir @ Arka Works: Yeah, I think people want to use it for things that they don't necessarily enjoy doing. I think where it starts to get slightly uncomfortable is where it's starting to be used for things that they do enjoy doing. And So the more creative stuff so and actually stuff lots of I look at is how do you use it creatively?
So when you're doing early stage design work or You're developing a design through how do you actually use it on a project say and so it does step into that realm? But what I find is it doesn't feel like it's taking stuff away It [00:05:00] actually feels like it's adding to your process and making it better.
Stephen Drew: I agree. Now, if anyone in the audience while we're going live does want to ask any questions, you can. Although, they need to be, we can't have anything controversial. I don't want to get banned from LinkedIn, but equally. And we will definitely try to answer any PG 13 questions as they come in.
Maybe what would be cool, Kier, because you have put the effort to do an awesome presentation, which is much more effort than I normally do. So it would be cool to see that. Maybe if I can bring it on the screen quickly, and if it's okay with you, you can just give us a little bit about yourself more expanded on the business and your ideas.
Keir @ Arka Works: Yeah, absolutely love to say Yeah, this is a presentation. I've been I gave in Oslo recently for a PropTech event and I'm actually being, I'm coming into practice a lot to talk about how they're going to start to transform the work that they're doing. And this is a useful this is a bird's eye view of how I see.
Things moving. So this is my working day in November last year. This is the work I was [00:06:00] doing at Morrison Company. So I was director on some pretty major projects, delivering offices, big housing health centers, education, buildings, all kinds of stuff. And then I I'd like to say that something felt like a meteor hit me on the next day, which was ChatGPT released its first version, and this was a really big moment, and it's hard to imagine it's still less than a year since that happened, but because so much has It has developed since then, and we were also testing a product called Delve, which is the icon on the right.
That's a generative master planning tool, which was made by Google and Yeah, so I was doing these two things at the same time And I put them both together and couldn't I thought I couldn't every time I was going back to projects I was just seeing things slightly differently like imagining straight away.
How would this happen in the future? And, this is a little, this was a photo I took when we were testing Delve in the office when you give it all of your parameters, it starts to run the algorithm and it unironically says, we're looking for the perfect designs, estimated time, five [00:07:00] minutes and I I think this is an interesting insight to how computer science may look at the sector of architecture. A sector which has never been something that they could target as a market because it's knowledge worker dominated. It's something that is full of intangibles and qualitative thinking. So how on earth could a computer do it?
But what I think we've done a lot of in the last few years is we've turned quality into metrics and KPIs and different quantities. And actually what that does is it's, it starts to create a sort of more algorithmic approach to overcoming site constraints. And so that's why tech companies think that they can provide great solutions. So this is how I like to think about the future. I think there's sort of two routes. There's anyone being able to pick up an AI tool and use it for architecture. Which I don't like the sound of or there's actually really well trained designers and placemakers using AI to enhance the work that they do.
And I think unless we get on board and start doing this stuff, we're going to be in the first category. [00:08:00] And that, I find that quite a sort of Depressing outcome potentially. I think if you've got really high, highly skilled people picking up these amazing tools and augmenting what they already do, that's like a superpower that there's no comparison, okay.
Between the two the designer, the person with good judgment is always going to do better. So that's really what the mission is of ArcaWorks is to augment practice. Improve the work of architects and I do that. We do that. It's just me at the moment. I'm in my second month. I work with architecture practices.
I do a lot of that and that's probably because I've come from architecture. And so I know quite a lot of architects. But I'm also working with tech startups, so there's a couple that I've got relationships with that I'll talk about today, and one of which I'm working for in a lot of detail called PreOptima, which is carbon estimation using generative tools.
And also starting now to talk to clients as well who, there's a real hunger from clients to actually pick these tools up. And do [00:09:00] early stage business planning and viability with them. So ArcaWorks is sitting in the middle of the stool, the three legs of the stool and trying to pick up learnings from one realm and transfer them across and trying to innovate on projects live, essentially.
And so the things that we're doing is transformation planning. I do a lot of workshops with practices If people have actual projects that they want to bring in expertise on I can combine the tools that, these new tools that maybe people are less familiar with actual project experience too, so I have a sense of what is and isn't useful.
And doing applied R& D essentially, so some people have been, wanting to get in touch to talk about ideas that they've had. that they want to build, they want to go for funding for, that's something we do with people too. And it's it's a creative consultancy. It's not it's, what's super interesting about this realm is it's computer science, but it's actually really creative.
And you've got this it's quite intuitive the skill set that you need. So these are all some of the experiments that I do and I [00:10:00] try and just share my learnings and everything that I do publicly. I'm very open. If people ask me how I've done stuff, I will explain and I'm happy to come to practices and talk about it.
It's not, I'm not keeping it within the castle walls as it were and trying to turn it into my own proprietary stuff. It's totally open. So these are some of the, a lot of the stuff I use is things I'm very interested in mid journey as a tool of kind of bringing references from random places and splicing them together.
I'm also using stable diffusion a lot for rendering straight from very early stage line work and massing. And this is a, so I'm doing things from like material palettes up to master planning at the moment. So this is a project that I've been doing with May actually. So this was a sort of applied R& D example where working with them on a master plan, which I'll show in more detail.
And then there's this, I won't go too into detail cause it gets a bit tacky, but essentially. We mix the terms, right? So everyone talks about generative AI or AI and they mean everything, but actually [00:11:00] that's actually quite, that's a different niche. So generative AI is more qualitative and creative.
Generative design is more about parametricism and deterministic controls and things that you've known about for a couple of decades, essentially. So these are all the tools that I've There are hundreds of tools coming, to be honest, but these are the ones that I think are relevant and useful.
And with many of these, I'm able to talk directly to the founding team and the development team and find out what's coming, why they're building what they're building, and have a sense of their roadmap as well. And they're giving me a lot of early access. To beta testing and things like that so I can actually get under the hood and find out what they can do.
This is a bit of fun, but tell me if you think i'm wrong so this is my take on the profession which is that All architects are on this sort of bell curve distribution and it runs from analog processes to digital and some of, some architects really digitally led and some are analog led.
So I [00:12:00] guess at Morrison Company in my work there, I was always trying to bridge between like handmade craft. We made a lot of models. We're very haptic designers. We were thinking always about materiality and trying to represent space through models and things like that. But we're also, Revit first on everything.
So all of our projects were, we did a lot of BIM level two stuff. So we were quite digital. So we were in the middle. And the way I see it is that the artisans live at the extremes. So you have like handmade craft led artisans, like Peter Zumter, an architect I love, sitting right at this analog end extreme.
And then on the digital extreme, you have digital artisans. So those are people who don't just do digital. It's, it is it's a digital led. Process it's very much digital native you're using Grasshopper right from the beginning and you're rendering everything and there might not be all that much touchy feely making stuff happening.
And the reason why I show this chart is I think in terms of disruption I think the [00:13:00] artisan business model is probably quite robust. You've got you've got usually quite like big name practices that they're reputable, they're famous. They probably got books written about them. Clients are going to want to hire them for pivotal projects.
I don't think that business model really changes that much because that's always going to be a need. So there, there might be some minor change there but I don't think much where I see big changes in bulk, the middle of the market, as it were, which is where the demands in the middle are very different.
The clients in the middle, and this is speaking from my own experience, they care a lot about standardization. They care a lot about. design work that sells either in rental form, if it's a building, a speculative office building or housing that's going to be sold or even affordable housing.
It's still focused on building stuff that, that is interesting, that will appeal to the market as it were. And it has to be like tightly cost controlled and it has to be readily buildable. And all those things are really well suited to this new technology. [00:14:00] And the reason I say that is I think architects need to be really clear about where our potential vulnerabilities are and why that means why we're potentially ripe for disruption.
We'll all have heard these things, I think. These are some of the bad things that clients say about architects. They say that we applied a bespoke approach to every site. Now this is almost something you would wear as a badge of honor at architecture school. And I always felt like this is no site is the same.
But for a client, that's annoying because actually they just want it. They don't want, they don't want a sort of a handmade watch every time. Then this idea that our fees go up when the price of something goes up that's the if a brief has changed late, it's costly. And it takes a long time that we have quite laggy data.
So it might take us three months to get you an answer on the best embodied carbon optimal approach for a retention scheme or something. Those are all things that frustrate clients and they're all things that. In the future, generative models and platforms will be able to do really well. And actually go one step further, which is to look at [00:15:00] myself as a sort of egocentric architect.
Stephen Drew: egocentric architect. My goodness. Come on then. Let's see what you got to say about yourself.
Keir @ Arka Works: so I think this might be something we all suffer from a little bit, but I always say we kind of design with our monograph in our heads, I think. So there's always a sense of this magnum opus that's building over the, over every project. Which is great. But at the same time, how is that serving the client on their project?
Do they even care about your sort of portfolio? How are you arriving at your design solution? Sometimes it's quite esoteric we might build like 50 models of something, and that's something that feels like it's the right way to approach something for me, that's how my brain works, but sometimes that doesn't I feel like a good fit for clients.
And then I guess we're stubborn about ideas and, sometimes we're not building, designing to cost perfectly. And we use like phraseology. That's a bit of juice, like Archie speak. These are all things that, that, that mean that maybe when a client asks for [00:16:00] something, it's not exactly heard.
So these are all things we have to be, I think we need to be self reflective about that and understand that those are in some ways strengths, but they are in some ways weaknesses. And then I'm just going to give you a quick look at where I see current use of cases, right? So on the practice, I think about all workers like either sitting on the practice or the project side.
On the practice side, these are things that I'm doing right now with these tools. So using them to write really good bid responses on fine tuned. Large language models to do report reviews. Reading really long reports and getting quick summaries before meetings. It's very good for that. You mentioned marketing collateral.
That's like very good. Taking long form content, turning it into short form doing like marketing plans for things processing data, the new models, the new model of GPT can ingest Excel sheets and really good, quite data from different sources and actually put it together into a single data frame, which is very cool.
And you can write code as well. One of the things I do is I'll produce tables and things in Chatty Beauty and then actually send them [00:17:00] back to Excel, or you can send them to different platforms using JavaScript. So there's this the things you can do with every model release are increasing.
And then on the project side, these are things I'm doing. Anything that's prototypical, like housing or office is very good for plan generation. You can do generative massing with it. Environmental analysis, areas, materials, takeoff rendering, image editing and even just use it as a creative partner sometimes.
I think it's quite good. Can be quite good for that. And then there's all these future things coming. Intranet smart intranets within every organization where all the policy documents have been ingested and you just query them with natural language. Fee estimation, you could provide, provide all of your historical fees and you could generate fees very quickly.
Very good for QMS, reviewing documents. There's all kinds of, yeah, building compliance is something that's, that'll be coming through. With the Building Safety Act, the list goes on. There's almost, it'd be quicker to write a list of areas that won't be touched by this these technologies, in my [00:18:00] opinion.
How does this all look on a project? This is a project I did with May like I said, it was a live research project. And this was, we did some options for 1500 home master plan, massing layouts, areas, environmental analysis, and carbon footprint. We did it all in about 10 days together. So for this, we were using a massing tool, a generative massing tool called Spacio.
And this tool is, I think of it like generative SketchUp. It's rather than drawing lines and walls, you're drawing whole buildings at a time, and then you're pushing and pulling whole buildings and whole facades. And the buildings that you're producing have been pre programmed. You program them to the criteria that you want, and then you experiment with those things.
So it's. It's very good for anything like residential, essentially, that's prototypical, and it's maintaining a perfect record of areas as you go, so that's something I've always struggled with on large master planning, and [00:19:00] I think it does a really good job of that. So this is the kind of level of detail that I might get for a massing option.
This was the third option we ran and this was just a screenshot raw out of Spacia. So it's a quite communist looking just a grid, I've set the grid, I've set the sill heights, I've set the cores and so on and I took that and then I actually drew straight on top of it. So often there's a, I try and interdisperse the generative processes with handmade, hand, handmade work or hand sketch work.
So this is sketched on top. And then this becomes a control net for my render. So the next image is a render that I've produced from that control net. So this is using stable diffusion. And here I'm able to experiment with things like public realm defensible space materiality of facades, blending facades in different places.
Character zoning all kinds of stuff I can play with different types of materiality. So I'm, and I'm doing all that straight from that sketch that I just showed. So there's a really interesting hybrid approach here, [00:20:00] which is it's very digital, but also very. Handmade as well
Stephen Drew: Beautiful.
Keir @ Arka Works: Further was, this is a tool called Parallelo.
It's built for automated residential plans. This is planning apartments and they are compliant. They're actually, it's a Norwegian tool, so they're compliant in Norway, but not the UK yet. And this is very good for maintaining really good areas for a project. So you can just turn the dial if you want more two beds or one bed.
So you just turn the dial and it will recalculate your plan. So the idea is it's giving you not a blank page at the end. Just start when you take this into your, into Revit or into CAD or into, or a hand drawing and you start designing on top, you're drawing, you're starting from prototypical plans that give you 50 percent or 60 percent of the progress.
That you would have made that you would have needed. And then you take it from there. So this wouldn't be, this is, you wouldn't use this for a very bespoke project, but something that needs repetition and has types. It's very good. this is environmental analysis. So the same.
Same project is taking it now into, I think project a product [00:21:00] called Forma, and here I'm running daylight assessment, sunlight, potential assessments live on that model. And I'm getting really quick feedback that certain things don't work and I need to adjust my design. I'm getting quick feedback on wind discomfort, risk.
This is, so this is an AI that's been trained on wind studies. And that's a process that it's a CFD analysis that takes hours, maybe even days. And usually you do it at the planning. You usually do it at the end of stage two or three, but here you can do it at stage zero or one when you really just got some massing.
And so you can actually design with it and actually change your design. So it's a very different way of working. So you make good decisions up front so that when you run your CFD analysis with an engineer or a sunlight potential analysis, you actually pass and you don't have to post rationalize why you failed or why you didn't get a great score which arguably is how we currently do things.
This is a tool called PreOptima, which is this is a company I'm working with. It's a UK based startup. [00:22:00] PreOptima are focused on carbon estimation and LCA at very early stages. So you go straight from a 3d model and the algorithm runs an engineering analysis on the building form, and it works out how many columns, floors, beams, substructure, and it sizes all of those elements.
In a way that's conservative so that when an engineer comes along and does a finite study You're gonna probably end up with less material because they'll engineer it in a more refined way But straight off the bat you can test like three different configurations on carbon and say this is a more This is a good sort of carbon option based on building form which has a huge impact on embodied carbon So the same, 10, 000 square meters in a different shape will have a completely different embodied carbon, even though it's technically the same amount of building.
So it's trying to make LCA very simple just by applying systems and assemblies. That would be typical early on in a project and getting your LCA running. [00:23:00] I guess the conclusion of this whole thing is we've gone from this linear process where you have these discrete gateways and I, each step is like a fire engineer or a client or a QS, someone that can say something's wrong, go back and you get, you go back two steps and then you get stuck in this infinite loop of testing and pain.
And. So the way I see it, this is my, and I'm quite, I'm optimistic about all this stuff, is that you could take out a lot of that complexity and you can just work in sprints. So you bring everyone into a huddle, you design really fast, and you get, you validate your idea very quickly. So you don't have to go through these like really arduous.
painful steps in order to validate that your idea was good in the first place. So it's a different kind of way of working.
Stephen Drew: Wow. Is that the end? Oh,
Keir @ Arka Works: That's it. That's the conclusion.
Stephen Drew: was good. I was engrossed. I was like, and then? And you're like, oh, that's it.
Keir @ Arka Works: What else do you want?
Stephen Drew: But yeah we've got to see where it goes from here. That's [00:24:00] the thing. I'll be honest, Keir, when you put that spreadsheet, not the spreadsheet, but like the chart showing all the different brands and names that are in these spaces.
It's actually quite a lot, isn't it? Cause I know Modulus and I know Co pretty well, but there's so many others as well. A quick question on that. Do you think all of them have a purpose or do you think a few of these? Pieces of software are competing to be the de facto one, like Midjourney and DALI.
What's your thoughts on all these different offerings?
Keir @ Arka Works: Yeah, it's a bit, it's a bit, so when I've talked about this before, people say this is a bit complicated. You're using five different things rather than one. There's, so there's different kind of approaches. So the, I, the version I like is that maybe the future is less centralized. You. Rather than having Autodesk, say produce Revit and AutoCAD, and you live completely within that ecosystem, you actually can pick up tools that are more purpose [00:25:00] built and specific, and then use them for what they're good for.
Now, I mentioned Autodesk, they've got the project the product called Forma, which I just showed, that is intended as a platform. Where other people connect, right? So a lot of the, a lot of the earlier products I showed, the intention is that they will connect into Forma and you will access them through one interface.
So that's quite interesting. So they're making a play for, we're going to be the platform for this new. Realm and everyone can come and play in our playground and you can connect things up easily for people there. For example, test fits got a really good car park planning tool, which isn't a big thing in the UK 'cause we try and take cars out of projects, but in the, in America it's a big deal.
And you can access that car parking tool from within that platform.
Stephen Drew: Got it.
Keir @ Arka Works: so I, but I, yeah, as much as possible, I think it would be very healthy to move towards. Lots of independent specialized companies and products coming through that [00:26:00] get really good in one niche. And then I think you'll find that they'll get really good at one thing and then they'll start to spread out and they'll start to try and do more and more. And that would be very interesting.
Stephen Drew: makes sense. It does feel very Autodesk esque to try and build the marketplace for this stuff. Autodesk, if you sponsor me in the future, I will take out that comment. Not that it was negative, but I will definitely be, I will be. I'm up for being in your marketplace and I'll tell you where my bank details are.
I'm joking, but it's, I agree. I think we need a lot of different things. To spark innovation in this space, it's like when VR started in video games and stuff, it was all actually the little indie games that pushed the platform, and then you'd get one big title or two, but that's where the innovation came from and I think really important.
Me and you talked earlier, before coming on here, a little bit about NFTs. Now, Kim, let's pretend I'm cynical now, right? I'm actually an optimist as well, but let's role play a second, [00:27:00] right? NFTs with the big thing and the Metaverse. I was almost driving people insane last year, Keva, talking about the Metaverse too much, right? I thought it was important, but what I mean is, that came,
Now. Me and you said we believe it's going to come back, however it went quieter. Some people were saying AI is a bit of a fad, I'm not interested. What's your raw thoughts on that? How would you answer back to that question if someone is saying, what a waste of time or it's all going to blow over kind of thing?
Hahaha. Heeheeheeheeheehee. Hahaha,
Keir @ Arka Works: was personal computer a waste of time? I think, and I, and also, by the way, I do think
Stephen Drew: I got a role play. What? Yeah. Go on. Go on.
Keir @ Arka Works: No, I can see that. And look, it is a bubble for sure. There's definitely a sense of euphoria around it and a sense of like hype and froth for sure. And you can see that reflected in like Nvidia stock and all of the other things happening in the [00:28:00] markets around it. And all of the tech companies trying to nudge each other out of the way, and everyone's got their acquisitions.
So all the big. Tech companies have all acquired some major sort of AI company, like Amazon just bought a big stake in Anthropic, Microsoft have OpenAI, so that they're all trying to own that space. So yeah, I think there's definitely a, there's like a bubble element to it, but I think you could probably say that about any transformational change is that it starts with enthusiasm, and it starts with kind of crazy hype and optimism, and then you have the Inflated peak of expectations and the trough of disillusionment and the whole hype curve will play out.
So no doubt in a year, or I don't know how long it will take, maybe six months, we'll be in the trough of disillusionment and then we'll pull it, it'll be pulled out from that. But I think if you zoom out from 10 years, I think in 10 years, things are going to look extremely different. I think most people, I think there's a lot of consensus about that.
I don't, and I don't think. Most people would put AI in the same bucket as NFTs, much as I like, actually, the technology [00:29:00] around NFTs, by the way I, yeah, I think it's a platform shift. So it's huge. It's. It's it's similar to changing the, how the work changed after we invented personal computing in my mind
Stephen Drew: I see it as CAD to BIM, and there was a few businesses that were reluctant to go to BIM and, and we are being pulled that way. I do think AI is going to have a big part of it. The only thing I would add to what you say It's where I do think people, I can understand some people being jaded is that I see a lot of tools aimed at businesses which claim to be AI or the genuinely not AI, so there's, I think that there is the AI market in public. Oh, we've got a chatbot, which is AI. And I'm like, dude, I know you're just pulling in chat GPT 3. 5 or whatever, and rebranded it. Oh. Before that, they used to, in different fields, be like, we have a new AI tool. And I'm like, that's not AI. But I do think there is proper [00:30:00] AI tools now.
And what you're, you spent a bit of time earlier talking about the real ones. Is that fair? Do you see some people trying to get on the AI bandwagon without AI or
with the real tools?
Keir @ Arka Works: I think that's why I have that, that, that chart, which was the whole spectrum, because I think there's. Yeah, people, things are being labeled as AI that aren't really, but they are still really useful and they are still generative. They're just using a predetermined set of criteria to get a known outcome.
So they're deterministic. And that's generative design. It's, so it sounds like a semantic issue, but it's important because when you're using something, you need to know how it's working. And also you might believe certain things are possible that aren't possible. So I think people have assumed that you can generate things
that you can generate plans or CAD details using AI. That's actually very hard to do. There is some really interesting research on plan generation, I think, as a project called Archigan. That [00:31:00] was done learning basically training models on architecture plans and generating things.
But it's very hard to take that and then go build buildable plans that are measured and have like vector lines and finite quantities in them. And actually when you think of it, when you have it, when you understand that it's a bit less scary, because really what's happened in my mind on the generative design stuff is you've got, I can't, Grasshopper is really difficult.
So your average architect isn't going to be using Grasshopper, but these tools have put a wrapper on them that basically mean that you can now get the sort of benefits of a Grasshopper like approach in practice. It's not super tailored and bespoke, but you can get some of the benefits and you can use.
You can then use generative. It's a generative sort of leverage that you add to your process. That's the shift. So you've, they've gone from being like R& D inside top architecture practices to you can buy a product or use a product that's put a really clean wrapper on it and you can access it through your browser and it's really easy.
That's, so that's, and that's really feels to [00:32:00] me like the last Basically since November, that stuff has all happened. So the generative stuff has been around for a while, and now it's getting bundled in with AI. And I think about it like two hemispheres of a brain that one is creative and one's rational.
And actually if you can use them together, it's a brain and it's really great. But you just one on their own, one on its own isn't necessarily as interesting.
Stephen Drew: It's very, it is that is a good way to look at it. I hadn't thought of it like that before. Now now we are going to mention in a bit where architectural practices can get in contact with you, because I do think it's important that you have people like yourselves, authorities in the sector or a researcher doing all this stuff.
Just before we go with that, though, I have Two kind of questions on that, but number one, okay, for maybe not the business owner or anything, say now there's someone who's an architect in practice, maybe they're a pot who is enthusiastic about this stuff or an architect, where would you encourage them to begin the [00:33:00] journey in AI so that, because you've done all this stuff, so just to avoid a few mistakes or where to look, do you have any quick pointers to put the working professional in the right direction?
Keir @ Arka Works: There's some good there's some good stuff online by Neil Leitch is a good author, like he's a, an academic working in the space. The Archigan project was done by I'm going to say his name wrong, but Stanislav Chilyu, I think is how you pronounce it. He's a founder at Rayon, I think now but he's just released a book which is very good.
So that's what we could be, would be a good place to start to get the sort of history and one of the things I do, so one, as I go and do workshops for practice, part of the workshop is just how has this technology come about and why is it happening now? And why is it different? And I think that's part of what they, those people have done to help progress the agenda.
Then there's just really, there are some really great, I have to say, like yourself, Stephen, I'm all over LinkedIn. It's a very good platform for idea sharing. And what I find is when you start putting out interest on certain [00:34:00] things come back. So people will reach out to you that are working in the space and you'll find things out that way.
Or you can just, I get a lot of my ideas just from things that people share openly on there. There's some really good people, if you're into mid journey or you're into stable diffusion or you're into generative tools like Spacio you can join communities that the sort of early adopters of those things and you learn an awful lot from that and I and I try and do the same thing.
So if I learn something new, I just put it out there or I'm happy to answer it. I think there's a generosity there. Like most of the people that I follow, so there's a guy called Omer from design. I'm going to get, I'm going to get this thing wrong. Design input, design. input. He's got great instruction on Stable Diffusion.
Yeah, there's there's Ismael I'm going to say his surname wrong, Silliet, I think. He's now gone to Foster's, but he puts a lot of stuff out there. So there's some really great voices that people should follow. Yeah,
Stephen Drew: Okay, cool. That's that is [00:35:00] really useful. Now the other last question for me, I was going to say before we, we you should tell everyone all the cool stuff that you're up to and where they can find that is if You are a student now with the knowledge that you have. If you were a student in this kind of point with all the stuff going on, part one or part two, what's your raw thoughts on that?
Is there, would you bring that into your projects or would you or where would you begin on that front? Because we're not talking about physical constraints,
Keir @ Arka Works: I think that is actually quite hard because I think it's all well and good using this stuff after you've done it the traditional way. But I think there's a slight, there's a potential for unlearning if you don't. Do the hard work of learning how something actually should work in the first place.
Stephen Drew: Yeah.
Keir @ Arka Works: So I, and I think you also need that to give yourself perspective and judgement on good and bad. So I do think people should, I think students should be learning about it, but I think they should be very wary about using it. Everyone should be very [00:36:00] thoughtful about the way they deploy it, right? I don't, for me, design is not prompting and getting an image of a building.
That's just not design. If you're, if there's some sort of collage process where you're providing a very speculative input and you're really leading that tool along, I think it could be a co pilot to help you in that process. But you, what the risky territory is I've got a good, I good idea.
And I let midjourney show me what it looks like, and
then that's my design
and I turn it into 3D model. I think that's slightly problematic because you're just, you're, that's just an aesthetic that's just in a veneer sort of image deep view of architecture. And obviously it's not that, it's a architecture's a very holistic synthesis of space, societal needs, structure.
Yeah. All kinds of real world constraints. So I would really encourage them to make sure they do the hard work. Of learning to be architects. And then once they feel like they've done that, to then use that knowledge to, to move into this space.
Stephen Drew: [00:37:00] Yeah. I think you're right. Any freshers who'd rock up to university on their first week or two and think that they can put a chat GPT image on the wall and get away with it with your tutor when you hangover, it ain't gonna work. I
Keir @ Arka Works: I think, I, what I would say is, I think stable diffusion is a really, because it's working from sketches and drawings, and it's not
generating designs, then fine, use that, go for it. And that's actually really hard I wouldn't be, I wouldn't, if I were a tutor, I'd be quite intrigued by that, but I, if people were giving me mid journey designs, I'd be a bit Eyebrow raised.
Stephen Drew: Yeah, fair enough. That makes sense. While we were talking, Red Mike says, Keir, what's the name of the book from Stanislaus
Keir @ Arka Works: Stanislav something.
Stephen Drew: Can you please post a link in the chat?
Keir @ Arka Works: I got it. Hang on. Hang
Stephen Drew: we'll get them all after. Oh, he's going to get the book! Do you want to do it after or before? It's Red Mike. Keir's running around. Look, here's a power one. Isn't it nice? Look at mine. Look at mine. Let's move up in the
Keir @ Arka Works: Here you go.[00:38:00]
Stephen Drew: There it is! There it
Keir @ Arka Works: a look.
Stephen Drew: Red Mike.
Keir @ Arka Works: Stanislav Chaliou. Chaliou. Terrible pronunciation. I think he's
Stephen Drew: No, very lovely. Red, Red Mike, I've met him in person. He's a really cool guy. Red Mike, get the book here. Thanks for running around that. While you were going in the background, I did say how much I like your apartment. Yeah, again, I'll stop gushing over your house and Now, let's bring it back to what you do.
You did mention in the presentation what you do. I like your stuff. You've got all that experience behind you. You've got a new venture. And so I'm going to bring it up, but tell people where they can find all your stuff.
Keir @ Arka Works: Thanks very much. Yeah. So you can find my website. That's aka. works, that's a r k a dot works. And yeah, it's all there pretty much. I put, I write a lot. So that's all up there. I'm writing for various publications but I try and post all of my writing here so you can find it. And. Yeah, so there's lots of research that's been shared here things that I'm looking at, [00:39:00] and links to all the other social media things that I try not to look at on my phone because I'm totally addicted to them I'm, I guess my main thing is LinkedIn.
I'm I think LinkedIn is the best for professional work and incredibly good for this new space. I am all over it.
Stephen Drew: yeah, it's the place to be. Me and you both. And for I'm like better called Saul in one way
without the greaseball angle, but I'm all shouting about careers and jobs and all this stuff. We need people. Champion in AI. And I think it is important, like you say, to have people in the industry like you, who are doing it, leading from the front and adjusting, isn't it?
And making mistakes and going, you know what? That was amazing for this. And now we've got this new thing out and this is what's happening. And now a water desk has come in. I gotta be careful what I say. All this stuff, right? So it's going to change the space. I really. I appreciate you sharing this.
Now, Kia last thing before you go, it's only fair if, I'm asking all these questions, you get to wallop one or two things at me. Have you got [00:40:00] any loaded bullets or guns you want to chuck my way?
Keir @ Arka Works: I'm not sure if it's, I don't know if it's a tough question I don't know how hard you're going to find it. I'm interested in obviously you're working a lot with people on their CVs and looking for career development. And I'm interested to know if you're seeing any demand for these tools for use, skills in these tools or training or, because I've had people ask me For example can I get a training certificate that I, from you?
Can I do training with you and can I get a certificate? And I'm not a school, I'm not a university or anything. I'm interested to know what people are saying about it on this from a CV perspective.
Stephen Drew: Wow. Good point. So years ago they used to be Revit testing, which used to be Knowledge Mart and stuff like that. And if you had done a test on it, you got a decent score. It could improve your chances with getting a job because. A recruiter like me in the past or someone else you go to an employer who doesn't understand Revit, chances are the director doesn't really use Revit, they just know they need it in the business.
And you go, here's my certificate where I got 83%, they go, wow, come on in, Steve, [00:41:00] you're going to do it. There is something in that certificate thing and architecture as a culture, we love certificates, we get CPDs, you get a certificate, all this stuff. So I do think there's space there.
The interesting bit where I'm quite controversial on is I've never been a huge fan of covering letters because I think that they hemorrhage time. And I think if there ever was an opportunity to use AI. in the job seeking process. A covering letter is the one, in my opinion, which
Keir @ Arka Works: Does it meet certain criteria and then you just don't read it?
Stephen Drew: yeah it's, I don't know if everyone reads a covering letter. They can spend a lot of time. You can spend ages customizing it. Some in architecture practices will love them. They will read them. They will go through it. Others, it's in the bin. They just look at the email, they look at the CV and the portfolio.
And so my, what I always say is that I think that cover and letter is a good place for you to utilize AI or do different interpretations. You still write your cover and letter per se, [00:42:00] but if there was anything to go through it, that is the one. Do I see? AI doing CVs and portfolios? No I think that basically a CV and portfolio is the most important document an architect can have, and in theory, Arcaworks is an extrapolation because it's a business, isn't it?
So it's manifesting in a digital and physical form, but people CV and portfolios like their own businesses, and it ain't gonna really cut it there, but a covering letter should be, A fluent summary of your experience and what you've done, which is a good place to utilize AI to direct that kind of thing.
Keir @ Arka Works: I, that's interesting. So I, yeah, I actually used to do it. Like I would be I'd be looking at all the emails at Monastica. I was often doing recruitment
Stephen Drew: you in the inbox? Okay, tell us about what you would be, how would you go about it as the person?
Keir @ Arka Works: You're looking for a certain criteria, obviously, but so we would filter by, we'd [00:43:00] put everyone into little folders of, okay, there's, what grade are they at basically?
And what were we looking for? And then I'd have a, so then we'd have a pre filter, which would be, I think we had 10 questions or something. It was like, do they meet all these criteria? And that's just to get into the shortlist that I will actually look at. So then I would look at, I would open the emails if, And what I want is covering letter, CV, sorry, covering letter and portfolio as two attachments.
I get confused if there's more than two. And I don't want links. I hate links to other websites. I hate download
Stephen Drew: are bad. Some people don't click links. And no one wants to download a 100 meg file from a Dropbox folder,
Keir @ Arka Works: to be, everything has to be under 10 meg total, right? Because it's got to take me no longer than about five seconds to open it. If it is longer, I will literally just move on to the next one, unfortunately.
Because the quantity is so high so that's the thing. And then in the covering letter, assuming you get to that point, you open the covering letter and [00:44:00] There's often clangers in there Dears are Hadid Architects. It's and you're Morris Company. That's no good.
Stephen Drew: Yeah. Yes, madam. Come on. Can you look at someone's name here?
Keir @ Arka Works: yeah, true. And then there needs to be some effort to, so often I would look at something, I'd be like, you've applied to the wrong practice. I don't know why this clearly doesn't fit. Like you're into something very different than what we do. That would be immediately just move on to the next one.
Unedited portfolios, if they're too long, like that's not a good sign. You've got, you've got to be that you want, they should have a sense of getting the information in the right order and it being digestible. And if they, if there's no editing, then that's not a great sign.
Stephen Drew: You, these are the stuff that I've said. I'm glad you pointed out one thing as well, that the reality is you're busy. People who are looking at CVs are busy. They will look for key things. And if you, if they don't see what they're looking for straight away, It's on to the next. It's sad, but true.
I use the analogy, sometimes it's a bit like Tinder, you're swiping. [00:45:00] And of course, I don't, but what I mean is that, you've got that moment to impress. And,
Keir @ Arka Works: I think, yeah, you might have only 30 seconds, honestly, or 60 seconds.
Stephen Drew: yes, people aren't spending 10 minutes on CVs, unless you'd like second stage interviews. Do we hire this person? Should I go through it like a tooth comb? They're marinating, umming and ahhing about it. Then it will go in details and, but initially it's, okay, who are they? What, Architectural Assistant, Revit, beautiful pictures.
Okay, let's do a bit more in this. And like you said, then you go, then it gets to the next level and the next level. And then you're right. Someone's made a decision in about two minutes if they're inviting you in or not.
Keir @ Arka Works: Yeah, exactly.
Stephen Drew: I don't think AI is going to replace that process though. Not yet.
Keir @ Arka Works: No, but I could certainly imagine it reading CVs and stuff and doing the pre filter thing that I described very easily, to be
Stephen Drew: Yeah, That, that's existed for a long time
Keir @ Arka Works: I haven't used that before. I didn't know that
Stephen Drew: no, it's more for big companies like Amazon and [00:46:00] stuff. But where I think as architects, it's going to be very difficult is those is what's called text parsing and looking at different keywords and stuff. It's a bit more nuanced in architecture. It's very visual.
If you see, if you scan the portfolio and just got the text out, it might not even make blooming sense because it's all the images, all this stuff. So I
Keir @ Arka Works: using like random language that doesn't make sense.
Stephen Drew: Yes, exactly. So we've made it difficult for Amazon's text parsing system, customer service person working or someone working in the Amazon warehouse, it will take that CV and it will work out if the person is right for the job.
It's just architecture. We haven't got there just yet,
Keir @ Arka Works: yeah, sure.
Stephen Drew: there you go. On that note, if you want to talk about boring recruitment stuff, I'm your person. However, if you want to talk about interesting where AI is moving in architecture practice. Keir is your gentleman. Once again, Keir, where can people find you [00:47:00] again?
Keir @ Arka Works: Yeah, just arcaworks, arca. works or usually LinkedIn, you can, you'll find me there in, as among other places, LinkedIn, I'd say.
Stephen Drew: Excellent. Thank you so much, Keir. I appreciate you being here. Stay in the stage one second while I say goodbye to the audience, because you know what? We've got more content coming. I don't know what it is yet because I haven't scheduled it, but I remember looking and thinking, oh, a few cool people have come in.
Not quite as cool as Keir. Maybe on the same level. I can't have favorites. I'm the guest. It's a privilege everyone's here, but thank you for tuning in. More content coming soon and see you all soon. Take care. Bye bye now.
Keir @ Arka Works: Bye bye.
Stephen Drew: Take care.