119: ‘Nothing Ever Goes as Planned’, with Stefana Parascho

119: ‘Nothing Ever Goes as Planned’, with Stefana Parascho

===

[00:00:00] Welcome to the TRXL podcast. I'm Evan Troxel. This is the podcast where I have conversations with guests from the architectural community and beyond to talk about the co-evolution of architecture and technology. 

In this episode, I welcome Stefana Parascho. Stefana is a researcher architect and educator whose work lies at the intersection of architecture, digital fabrication, and computational design. 

She is currently an assistant professor at EPFL where she founded the lab for creative computation. Also known as CRCL through her research. She has explored multiagent fabrication methods and the relationship to architecture. 

Her current research focuses on human robot, collaborative processes and the relationship between robotic construction and the built environment. Her goal is to strengthen the interdisciplinary nature of the field by increasing accessibility of digital tools And connecting technical research with societal [00:01:00] aspects. 

Before joining EPFL, Stefana was an assistant professor at Princeton university, where she led the CREATELAB Princeton. She completed her doctorate in 2019 at ETH Zurich Gramazio Kohler research Previously, she received her diploma in architectural engineering from the university of Stuttgart and worked with DesignToProduction Stuttgart and Knippers Helbig Advanced E ngineering. In this episode, we discuss the increasing accessibility of robotics in architectural education and industry, as well as the need to consider the longterm impact and trajectory of the field. In her role as professor she highlights the excitement and formative experience for students to participate in the ideation and fabrication of precise and complex structures We also talk about her background in architecture and her experience with computational design robotics And her current work with multi-robotic assemblies Ultimately Stefana envisions a future where robots are working alongside us as coworkers with the ability to execute smaller tasks and communicate with humans in [00:02:00] a more intuitive way so without further ado i bring you my conversation with Stefana Parascho 

Evan Troxel: Stefana, welcome to the podcast. Great to have you.

Stefana Parascho: Thanks for having me.

Evan Troxel: Can you give us an origin story of where you're coming from and bit of your backstory up to where you're working now.

Stefana Parascho: Sure. So, I am educated as an architect. I actually started my studies in my home country in Romania, where I studied for two years. after which I moved to Stuttgart to Germany. And I finished my studies there. So I got my, what counts as a master's degree, nowadays.

used to be a diploma degree. I'm. that old, that things have changed in the meantime, . but I got that from Stuttgart, so from Germany. and that is where I first got in touch with computation, [00:03:00] computational design, robotics, any sort of new digital tools that were up and coming in, architecture, and in the design scene.

I studied there, with Professors EZ and Young Nippers, and I think I, overlapped I wanna say, at the right time. For me at least, it was the right time because really when the icd the institute, computational design when, got founded. And it started back in Stuttgart.

That's when I was a student there. So, I got to be part of the first studios, the first courses really, uh, it was all super duper new to me and everyone around the school. it really opened up a new niche and, an interesting. corner of an interesting new field that I wasn't fully aware of beforehand doing my architecture studies.

And it definitely, got me hooked. so I finished my architectural degree there in 2012. and then I actually, well, I worked a little bit in practice, for a few years in [00:04:00] engineering offices and in, consultancies, geometry consultancies for architects, so more on the software side, programming, solving geometric issues.

And then I went to Zurich to do my PhD. So I did my PhD, uh, at et h Zurich with grammar research. they are really, and still are really focused and specialized on robotic construction techniques. So, that's kind of where I've focused or narrowed it down on robotic assemblies. , in my work specifically multi robotic assemblies.

and then I finished my PhD in 2019 and I was at Princeton University for three years as an assistant professor, I was teaching and, and doing some research there. And then, in, now it's been a year that I came back to Switzerland. I, uh, moved Lozan in the French speaking part of Switzerland.

And I am an assistant professor at E P F L, the architecture department. still doing what I do best. working with robots, so focusing [00:05:00] on robotic construction methods. we're specifically also looking at human robot interaction, multi-agent systems. where we can employ robots for design and construction and what the relationship, what the human, what designers, with builders is and, where we can take this technology from now on.

That's basically what I do right now.

Evan Troxel: Nice

There's a lot there. I'm wondering, you know, with robotics in academic scenarios for architecture, can you paint a picture of what kinds of robots you're using? You're talking about multi. Robotic assemblies. I mean, from that word structure alone, it's like a concert of robots working together to do something.

Right. And so maybe you can just give the audience an idea and me of what that actually looks like.

Stefana Parascho: Sure. I wanna say in architecture in we mostly work with industrial robots, for different reasons. One being because they're very widely accessible. They're easy for us to get. they're, I wanna say [00:06:00] easy to use, but of course, they're not like, super intuitive to use for anyone.

You need a little bit of, skill buildup, but they're not the most complex machines out there. They're actually just. , a ton of steel and some motors. they don't have a lot of intelligence, but they, yeah, they are accessible for us. They're versatile. basically you can put whatever tool you want on top of them, uh, and then make them do what you wanna do.

but the other reason also is because we're not roboticists ourselves, so we don't really develop our own machines, not to exclude that. sometimes we hack machines we try to come up with, new solutions. but we typically rely on robots that are somewhat available that we can get easily and we can start working with.

So that's why we mostly work with industrial robots. but we are always, I say we , I generalize. For most people in my field, um, I do mean me, but, I think we, we get easily excited about all types of robots. So, uh, it's, it's also really nice to work with other types of, machines, but we need that input [00:07:00] say, people who really develop the, the machines.

then, yeah, when I do say multi, multi robotic assembly, uh, it sounds like many, many robots. Uh, it's always good to start small and increase complexity over time. So I mostly work with two robots. That's what I call multi. but looking into the reason for me why I didn't stick to say one machine coming up with a process.

But I thought about several of them, is that they can start taking on different tasks, different roles. They can start supporting each other, and it opens up a whole lot of possibilities that goes beyond like just one plus one robot equals two, but all of a sudden you, you can do a lot more complex tasks and you can really think about what their roles are.

how can those change, what are Benefits? What are things that you can do with two that you can't do with one? 

Evan Troxel: right. I'm always remodeling something in my house and oftentimes I find myself alone holding something up over my head with, or maybe using my head [00:08:00] to hold something up I'm drilling with my right hand. And I don't know, studying myself with my left hand. And I can only imagine the robots do so much more just by adding one more arm.

Right. and that idea of working together to accomplish, I don't know, bigger things, better things, whatever, is, it seems obvious, right? but I can only imagine the coordination that has to happen between those two so that they don't clash. It it becomes kind of an art in itself of just programming those movements or, or sensing those movements in real time.

I don't know, may, maybe you can kind of explain how that works, because I can only imagine the expensive mistakes that could happen, with clashes in robotics.

Stefana Parascho: Yeah. that's a good point. you found the perfect example, of what one can do with several robots, cuz that's literally what we did. So, I used robots, um, to allow one to hold and support. a structure and the other one to, to bring a new element so they will alternate.

And you always have one robot that acts as a support. it's pretty much like what [00:09:00] you were describing of using your head, but instead get a robot. but it's true. Yeah. So it comes with many, many challenges. It sounded relatively straightforward at first, but then you get into it, and bump into, yeah, I wanna say challenges that were also, also because, Of myself being an architect.

I mean, I don't wanna, blame it or generalize that architects don't have the skills and knowledge of roboticists, but we, we don't, I mean, we just have a different, uh, background. So that's a lot of things to, learn on the go and figure out solutions without having that strong foundation in robotics.

So, the main problem really was avoiding collisions between robots, between the robot and the structure that is changing, uh, at all times and figuring out the robotic movements. Yeah, I, I, I don't wanna get too technical or I dunno how technical you would like me to get . I can, but like, the main issue with an industrial arm is that it has typically [00:10:00] six joints.

They're rotational joints that are linked, one after the other. the issue is if you want the tool to go from A to B, you can tell it go in a straight line from A to B, but you have absolutely no idea what the body of the robot will do to reach that point. So it kind of has complete freedom. Or the controller, the robot controller typically calculates that movement.

and you don't really know where the elbow of the robot goes, where a certain, yeah, whatever part of it will go and what it will hit. So just. Imagining and thinking those movements through the way that we do as humans in a three-dimensional space, where it's just like, oh, I wanna go from here to here.

It doesn't really work with robots. You just don't know. It's, it's impossible to intuitively tell how it will move. So, um, that turned into a challenge in itself. I did, I, I collaborated with colleagues also within architecture. we worked a little bit with the [00:11:00] roboticist, I wanna say at the very beginning, to bring in algorithms that exist in robotics, uh, for path planning.

So those, it's nothing new that had to be infected, but we don't have those in our own environment. So in c a d or in any sort of, Modeling environment that we would use to, to model and test the things that we wanna build. don't have access to those algorithms. So a big part of the work was connecting to these other tools that are out there and being able to use them to, in that case, to compute paths that are collision free.

That can go, from where the robot is to where I want it to be, without knocking anything over. none of it is foolproof, obviously, like nothing is, So they always are based on some form of approximation or on a certain resolution at which you calculate the robot's position.

So actually a lot of. avoiding collision was also just staring at the robot, watching [00:12:00] it move, going from A to B and hoping that, you, that I react quickly enough in case it is gonna knock something over. Um,

Evan Troxel: can hit the The kill, switch the button to stop it right? Immediately. Yeah.

Stefana Parascho: I mean, we, work with it in, manual mode, meaning we don't even need a, kill switch or an emergency stop, but you constantly have to keep a button half pressed, sounds very uncomfortable and is, um, just so in case anything goes wrong.

You just let go of that 

Evan Troxel: just let go. That's easier to let go than to push. Yeah.

Stefana Parascho: Yeah, exactly. Yeah, exactly. . But that means you're cramping, you have your hand constantly stuck in this in between position for days in a row.

Evan Troxel: It sounds like the perfect job for a robot. Yeah.

Stefana Parascho: Yeah, exactly. I know . I know. And, not official talk here, but, students have come up with very creative solutions to not having to hold their hand on that button, but

Evan Troxel: Well, and I mean, what's,

Stefana Parascho: not allowed 

Evan Troxel: can happen if [00:13:00] somebody actually gets too tired and they let go? The robot just stops and Okay. Like, it's okay. Right. Yeah. 

So if they, if they're cramping up and they need a break, can just let go and,

Stefana Parascho: Yeah. Yeah. No,

Evan Troxel: for a little bit.

Stefana Parascho: exactly. Can always take a break. true.

Evan Troxel: So, with your, with your work, can you explain the types of conditions that, that exploring for the work that the robots are doing? Like what kind of construction? You talk about robots working in construction. What are the things that are going on? Because I think back to. the, early two thousands, 2000 tens when every, pavilion was out there that was being assembled by, you know, robots using, I forget, it was like carbon fiber, right.

And it would cure the air. and every pavilion was kind of this, shape of it was completely defined by the scope of the robot's movements, right. It was like, can't collide with anything. And the sweep of the roof arc was like what the robot could reach, at least, you know? And, I'm sure there was, there obviously some design intent going on in there [00:14:00] too, but it was working with a heavy set of constraints to

create the forms that they were creating with those.

And I'm sure a lot of that was done at ITG Stuttgart. Right. And it, that, that work was being figured out there in. in real time and even in the public, which I think was kind of cool, like doing it out on display and having these pavilions built and, out there for people to see. But I'm just thinking about now when, you add one more robot to that situation, the forms that it can do are completely different.

So I'm just wondering like, what are the types of ways in which you're using the robots to help or aid with construction?

Stefana Parascho: Yeah. mean, pointed it out very, clearly and precisely how things went down. Uh, also that was during the time that I was at Stuttgart myself, so I'm very familiar with, those beginnings and the pavillions that, we all built back there. and, it's. Part of how my work went too.

So, I generally don't work with a specific material. It's [00:15:00] not that I have a fascination for one specific material that I've been, studying in depth and a really, you know, find it in all my work. But I'm a lot more fascinated by the process itself and figuring out what robots can do for all sorts of different processes and materials.

So I did work in my PhD, I worked with, frame, like structures. I call them space frame, like, because technically they don't fulfill the definition of space frames where all the bars meet in one point in a node. So anyways, it was a whole conversation at my PhD, how to call them, but space frame, like structures made of metal rods, with the intent of, generating, geometries with a lot less regular Arrangements of the members. So members could have different dimensions, different orientations, different positions. You can have a lot, numbers of elements come together in one node and all of that. So basically to free up the possibility of design there. And this, I think, [00:16:00] kind of lies or lied at the.

Base of most explorations that happened in those first 10 years or something of, robotics in architecture and design. was a lot of, excitement about we can build geometries that are not buildable without those machines or that are really difficult to build by hand. So these machines allow us to design and to build completely new things.

So a lot of the explorations that I did, went in that direction too. I do think at this point, right now as I'm leading a lab and I have students, PhD students, working here, and we're constantly discussing and wondering where is this going? Like it's been, the field has now existed for quite, some years, I wanna say almost 20 years, quite clearly existed before.

Not that before, there was nothing being done with robots, but like really established, a field of robotics and architecture. it is a bit of a question, okay, we can build all sorts of pavilions. they allow more [00:17:00] formal exploration than without these machines, but they come with their very, very intense set of constraints, as you pointed out.

so like where, where are we really going? What. we're trying to look into is, ma making processes more adaptable, making them, more accessible, easier to use, easier to implement. be able to, work with them just in a, in a more intuitive way, in a more natural way.

and keeping that, connection between the human and the machine. So a lot of, work that was done, it. required a separation of human and machine. You had to define the geometry very precisely. You then figured out all the paths and how it had to move, and then you execute it. Of course, there's humans on site and nothing ever goes as planned, but the ideal would've been, oh, the robot does it perfectly without, the need for humans.

I think that works really well for very special cases of construction and [00:18:00] for those pavilions, yeah, it works really well. do wonder what it takes to see robots being used a lot a lot more ease and, Yeah, simpler. Just, just an easier, easier use of robots in construction, on site, in wherever, wherever you need them.

I think it's this connection with the human that one can work with them together. One can kind of almost decide on the spot, Hey, can you help me with this? Uh, I need to screw this beam in here. Can you hold it here? allowing some form of communication and of interaction and adaptability without needing to predefine every single move that the machine does.

So that's a little bit of things that, we're doing here or discussing here. it's a new lab, so we're setting up a lot of things. There's more talk than results at this point. but these are really the, the questions that we're, asking ourselves. Yeah.

Evan Troxel: It seems like, the idea is you want to get it to a point where it's like using a [00:19:00] printer, right? because that's how, like all of these technologies that have led to this point. Like that's what's happened. saw that happen with laser cutters. We saw that happen with 3D printers where you feed it, the desired output, and it performs that, and you have to think less and less over time of all of the steps that it takes to get the motors to do the thing that you want, and setting the power level for the laser and it to do this before that, because if you don't, it's gonna fall through and it's gonna burn the model or whatever.

all of those things are the things that slowly get removed from the system thinking way back to the idea of postscript and laser printers, right. And desktop publishing. It was like, before that, it was such a crazy chore to go through to set up your document to do that.

And then finally got to the point where I could just like design it in Cork Express or PageMaker back in the day. Right. And it would actually print out what I designed without any knowledge of all of the systems it took to actually do that. Right. And so with, with robots, [00:20:00] that's the goal, right? well, you

tell 

Stefana Parascho: I wanna say yes, but not

yes, in terms of control, in terms of telling it what to do, ideally there's an easy way of telling it what to do and it really understands and then just performs it. But the what to do I think is very different between robots and most other tools, we've had before because robots don't have a specific purpose.

They don't have one tool, one action that they do that they're made to do. they're just an arm, right? So the, things that it can do, it's basically an infinite, set of possibilities. So I think there's another step, of how do we translate, or how do we communicate? How do we first figure out what would be a good thing to do, and how do we communicate that to the machine?

So for me, the difference would be rather than designing something and then telling her about, Hey, make this happen for me, I think it can be [00:21:00] broken down into a lot smaller tasks and then you can really work with the machine, almost like you would work with a coworker, although the machine has different capabilities.

But the ideal scenario would be you have your little helper robot with you, and you can kind of ask it for small tasks, small things of, okay, what do we need to do now? We need to screw this. Or what do we need to do? We need to figure out the right sequence of how to assemble all of these things together, which is not something that a human can intuitively, very easily do, but.

Hopefully, if that can be translated into a task for the robot or the computer behind the robot, like those tasks can go from screw in a screw to plan something for me. But I think that flexibility and versatility, if, if we could a way to communicate and to control that so there's not a fixed set of what is an appropriate task for a robot?

Is it a big task? Is it a small task, but that it can somehow have flexibility in between? [00:22:00] That would be pretty awesome and really complicated.

Evan Troxel: I 

want to get to the part where we talk about contractors in this equation, but I don't want to go 

there yet because what, just, went off in my brain when you were talking about this is, is the rise of Chat GPT and, and language as a driver, right? as like this great UI for,

Stefana Parascho: Yeah.

Evan Troxel: for driving whatever the computer, right?

For driving the computer. 

And so 

this. with robotics, where do we stand? Right. I know this is like what we would say early days for, for this. Right. But are you really excited with the developments in, this AI side of, you know, the chatbots and what it can do to summarize information or break a complex thing down into actionable steps?

I would assume that there's some pretty exciting application for that with robotics because the UI is something that everybody can understand potentially.

Stefana Parascho: Yeah. I am, I am really excited about what's going on there. I do think, uh, [00:23:00] it's like you said, at this point, it's really the very beginnings. the most useful thing that I've seen so far with, anything. Chat GPT related is, code, so actually not having to write code yourself, which already.

Burns down an accessibility barrier that was huge, for anything. And like that I can see happening quite soon for robots in terms of how do I program this particular machine?

Evan Troxel: right,

Stefana Parascho: The difficulty still remains that there's so many levels of, uh, what one needs to define before even getting to the point of how do I translate that into code?

So yeah, what should my robot robot do? what is its role in the first place? do I then translate that into, is it the movement? Is it, an action of the tool? Say open a gripper, close a gripper, or do something, that, there's so many steps before actually [00:24:00] developing that code.

So I think it's just, it's just a matter of layering. Layers of communication until you, you can break it down into that. I mean, I am very curious to see if and when that might become possible through something like a natural language processor. yeah, it's a bit hard, like whenever we think about this, the first step is define things very specifically, define the actions that the robot can do.

And like, rid of that step right now feels very, difficult. Unrealistic to say, like, I wouldn't know how, how, what would we base it on? But I don't wanna say it's impossible. Like, 

Evan Troxel: Yeah. It's in, it's incredible to see how fast things have 

evolved with even just the G P T stuff. And so I can imagine that your gears are really turning in your brains over there of, how it applies and, and how it changes approach. Because to 

me, that's where these big step changes happen, right?

Is, is actually in the approach, not even [00:25:00] necessarily the 

execution, but it allows you to think differently about how you approach the problem solving. Because now, like you just said, I don't even have to learn to code. I can ask it to write the code for me. And that removes a huge step of the process, which completely changes how and who could approach solving the problem.

Because before, that was something you had to do. It was a requirement, and now 

it's like, well, okay. , 

you don't have to do that anymore. Right. now the computer is doing that part for you. Again, you're gonna double check it. You, you have to have some knowledge of it so you can make sure, 

because again, mistakes are gonna be expensive here, but this is getting solved over time and this is a step on that new path which wasn't previously there.

And that is, it's gotta be really exciting.

Stefana Parascho: Now it is, it's always a better question of, maybe trust or mistrust. 

Uh, and a question of, okay, if everybody has access or theoretically if anybody could have access to telling [00:26:00] a robot what to do, will this lead to something useful or will it just lead to a lot, of random things happening because people don't think anymore about the foundations of what can a robot do?

Where, where does it make sense? what do I need? Yeah, like you said, you need some knowledge still to turn it into something useful. maybe it's a little bit, at least in my head, it's a bit comparable to. . what happened with parametric design when Grasshopper came out is like all of a sudden anyone can move a slider and make all sorts of geometries, but how much of that is in the end really useful?

kind of gets, yeah. Diluted. And you're like, 

Evan Troxel: Well, everybody, everybody's got a vite before noise pattern nowadays, 

right? So, 

we went through that phase because of what Grasshopper enabled people to be able to do. 

but I do also think that there's this, mean, you're, you're in an, a academic setting and you, said, you kind of distill this down to this.

you analyze. every [00:27:00] little aspect of it. And then I think of my kids who don't know what a file system is, they don't even use email. Right. And so you think about using computers and how fundamental it is to like know where to put your files and how to organize them. Why not so that you need 'em now, but because you're gonna need 'em in three years from now, and where do you find it?

And that just doesn't even exist anymore for kids. It's like, 

what? What's a file? what? do you mean? Like, I don't, 

it's just in my app. I don't need to think about that. and it's the same way with, like Google search. Go out there and start searching for something. And the, before Google, it was all directory based, right?

And it was organized in a hierarchy. 

We had Yahoo, We had Alta Vista, and it was, we're doing a lot of nostalgic things in this podcast episode. , it's fu it's funny, I didn't expect that to happen, but, but it's like everything was a directory and you would drill down and drill down and drill down. It was very much based on that 

file structure, that hierarchy. And, and again, like that just doesn't exist anymore. And now with ChatGPT,

you 

don't even need to [00:28:00] go to Google if you're looking for, an answer. let's just disregard the part where the answer could be wrong right now, but

it, 

could also be wrong on Google, right? The, the right answer could be on page four, not on page one.

So it's really interesting to me because we still kind of hold these assumptions that we need this, underlying structure. And what if, you know, anybody could do this nowadays and it's, Life goes on, right? just, it does continue. It is so interesting to me to watch that happen and acknowledge that that's happening, even though like, it's not how I learned and it's not how I would even tell someone else how to approach learning to do that thing because don't know this new way, I know this old way.

so it's, it's fascinating to me that not only does this open up a new approach, but it, it's like the old approach is almost like non-existent anymore. and it's not to say that, that there isn't a gradient there, there absolutely is. I, I think it's, 

it's all fascinating to kind of watch from the sidelines, but people [00:29:00] who are embedded in the 20 years of robotics research they're not gonna think about solving the problem the same way.

Somebody who is going through school now with fresh eyes, with a ChatGPT at their fingertips, like that is a completely different scenario. I think it's fascinating.

Stefana Parascho: No, no. I, I, I totally agree. that's what keeps me really excited about it I'm not the biggest fearful critic of it as I, I see some of my colleagues being, but I am really, really excited just cuz it's, new and it's different and what will it do, what we, with the way that we used to work, It's super nice to see

Evan Troxel: Well, and you're at the forefront of, of where those sparks happen. I think that, like you said, you're so excited about it and, to watch students, you know, 

ignite these ideas and to ask questions like they're five years old because it's all 

brand new to them. It's gotta be such a, a fun place to practice.

Stefana Parascho: Yeah, I did, I mean, last, last year, no, not even a year ago, but maybe like eight months ago, I was teaching my students how [00:30:00] to program in Python,

Evan Troxel: Mm-hmm.

Stefana Parascho: now I'm wondering if I need to still do this , if 

they can. 

Evan Troxel: Exactly.

Stefana Parascho: Literally, it's 

Evan Troxel: are, you are literally following in the footsteps of every architectural program out there that used to teach how to use tools. Right. It's like 

I used to do this, I used to teach Form Z and then I taught acad, and then I taught Revit, and then the schools were like, forget this.

We can't teach, we don't have time to 

teach these to the students anymore. There's so many resources online, they can just figure it out. So they actually just offload that responsibility to the students themselves. And now you're doing that because you can, right? It's like, what could we fill our, most important.

Subjects in time with if we don't have to fill it with that anymore. Right. I think 

that that is a very interesting thing to stay a top of. Yeah.

Stefana Parascho: No, definitely

Evan Troxel: So on the contractor side of things, I'm wondering how, and if, well, let's start with if the, is there any coordination or collaboration with the build [00:31:00] side of things? I think it's so interesting that all of the people, and maybe it's just because this podcast focuses on architecture, but that all of this is coming from the architecture side.

subjects that we've talked about with robotics in relation to architecture and construction all come from people who started out as architects. 

And so 

can you tell us, like what's going on on that side? Because I imagine that. . There's also some fear there, like that's coming to take our jobs.

but also there's probably some huge potential there. I know that there's some uptake of robotics, like especially on the 3D printing side in construction, right? 

As far as like being developed from the contractor side, to deliver product. So maybe you can talk about that, that side of things, because you're in an architecture school, but, is there exposure with students to the construction side too, it's like, you can imagine designs, but when you actually have to build designs, that is something else.

And then there is a wealth of knowledge on the construction side that architecture side doesn't necessarily have. so [00:32:00] what is that collaboration like?

Stefana Parascho: Yeah. Ooh, I have a lot of thoughts on this . I have to structure them in my head, but I mean, maybe to begin with, I think I'm in a very luxurious academic situation, where, am really free of the, say, demands of the industry, and I can develop things that are interesting to me for a, of course, there's a reason behind things.

It's not just that I wake up in the morning and I find this interesting. it does lead to the fact that. I think my research is a little bit more on the fundamental side, it's far away from what actual fundamental research is. Like. We're not doing mathematics, solving some sort of, I don't know, questions within.

Numbers that we don't know where to apply it. it still has an application. but I, I get really excited about the possibility of what if we could do this? What if we could do that? and because I am in my luxurious academic position, I can be very [00:33:00] critical of the industry. So rather than catering to their needs, I can be okay.

The industry has a lot of issues. it is maybe not moving in their whatever. one would consider the right direction, say, with regards to dealing, uh, with the climate crisis, dealing with social issues, workers' conditions, et cetera, et cetera. I can be a little bit on my high horse and be like, you're all doing it wrong.

Um, I can criticize you and I'm not gonna offer you solutions for what you want, which is to be more efficient, more cost efficient, faster to build more. Build cheaper. you know, make more money. So that's where it's always a better question, uh, where we engage with industry.

In, in what way we that knowledge out without being the ones catering to maybe what we see as problematic ourselves. I try to do that. I mean, it's not that I don't work with the industry at all. I do [00:34:00] look at what's going on and what could be improved, but that's, for example, one thought where this close collaboration between workers and machines comes from.

It's less about making it more efficient, making it cheaper, making it faster, but maybe about making it safer. If you have a machine that can, take over tasks that are really not. great for workers' health, that might be a bit dangerous. I mean, are really big statements.

Don't take them to literally, but maybe it can make it more equitable. can have people work on construction sites that maybe don't have the, say, Physical capabilities of a one meter 80 tall, strong man. you have a machine that can make up for this. and yeah, and maybe it can make it more creative, like less, uh, that everything needs to be predefined and you execute.

you are a small part. You execute a task that's been assigned. But if you have a. Partner, be it a robotic partner that can kind of take over certain [00:35:00] parts, that, that gives you more adaptability that allows you to react and just be like, oh, let's try this, let's try that. it's somehow how I would imagine this, uh, a bright, nice future of, uh, what could this be useful for, uh, in terms of industry?

On the other hand, it's like you said, there is so much skill and knowledge on construction sites that we don't have. And our architects, in school only learn so much you can only gain so much experience, an internship or something on site. but we're actually talking a lot about it right now because we, we look at human robot collaborative processes and we're wondering where where does it make sense that a human with their skill and expertise comes in and aids the process? And we realize we can only guess cause we're not the ones with the great skill and expertise in construction in know the things we try out in our lab, but, we're not carpenters.

we have worked for many, many years and have gained very specific skills or we're [00:36:00] not, specialists and, yeah, and anything working with steel, welding, et cetera. So, uh, is a bit of point where we're thinking about, uh, do, do we need to bring in people and see how they work with the material to see where would it make sense for a robot to add?

Or do we need to go on construction sites and become those experts? I don't know. Those are questions on how to push this research forward. there, definitely, need to link it more and. To understand basically the skills and knowledge that is out there, the expertise that exists on construction sites directly,

Evan Troxel: the risk 

avoidance that architects traditionally have, you know, around means and methods, and you're breaking down those barriers in school, right? By 

having the students. and your research teams actually design and manufacture, build, fabricate these structures and I think. , that needs to happen more, not less, but I don't 

know how that actually happens because they're, the separation is so clear, right?

Between what architects [00:37:00] and 

design teams do and what they're quote unquote allowed to do, right? insurance-wise. And you start to think about 

all the layers of stuff that is in between design and build. and obviously some, companies have done more in the realm of design build than others, but it's still like, just looking at the big picture, there's a very clear kind of barrier there.

And yet, like the opportunities that you're starting to identify or that you have been identifying in what's possible with the built environment, with the use of these tools is so obviously exciting about, you know, for, and I would assume exciting for both sides because I can see why some people see this as threatening.

And on the other hand, it's like there's so much opportunity of things that we've never been able to do before that, like you said, now that this thing can do this task for me, what if Right? It becomes the question, well, let's find out. And I think that's obviously, that's a huge, [00:38:00] exciting aspect of the environment that you're working in every day.

But getting that out onto the construction site, I mean let's just talk about this for a second. When students get into the industry, if that's where they go, if they don't stay in academia, are the ones who are listened to the least, right? And they have the exposure and they have these ideas and they understand the opportunity, the possibility, because they've had this experience and they're going out into a workforce where that does not exist at all, and nobody wants to hear about it.

right? Not nobody, but I'm totally generalizing, but for the most part, and so like, how do we get more, exposure maybe through podcasts like this? I don't know. But have you thought about that? it seems to me like you have, a potential audience because you have a stage where you're teaching to bring in people from industry to expose them to it, on both sides, on the construction side, but 

also on the architecture side and on the engineering side to say, used to not be possible.

Now this is possible and you need to know about it because there's graduates coming out [00:39:00] of our program who know how to do this, are you doing anything like that?

Stefana Parascho: I mean, any collaboration with industry is a bit targeted to that. I wanna say, anywhere where we, manage bring in a company, whether they're. a company of processing a certain material or construction, or whether it is more on the, architecture, engineering side, I think the goals for like the biggest win for us is if exactly that.

If they understand what's happening and how possibilities are changing and what sort of knowledge, we now have and the students that come out of here, it's a big hidden win because any collaboration is focused around the output and what are we developing and who brings in the money and what's, where is this going?

But I think there's this very particular, big wind there. it's true, it's, difficult, we educate the next generation and they're only at the beginning and maybe in. 20 years from now, they will be [00:40:00] the ones calling the shots and be able to apply a lot of this. but it does feel like a very slow process.

so yeah, no, I don't have the right, I don't have the magical solution to that. Um, that connection to industry is important for sure.

Evan Troxel: It is 

interesting to think about the students who are gonna be in those positions in 20 years and how much robotics will 

have changed in those 20 years that they haven't been 

using robotics, because

Stefana Parascho: It's true.

Evan Troxel: is kind of a reality. And so it's just this really slow treadmill , right? that we're on of 

advancement. 

and there's a lot of waste in that. It's unfortunate, but I would imagine that there are some firms and some outfits out there who totally see the value in being a part of this now, because it can differentiate them a world that is heading more and more towards commoditized architecture and, and services, but also output.

And so 

like 

you said earlier, you, you're in this luxurious position of being in academia and you get to be like the pundit of architecture. Guess what? Yeah. [00:41:00] This is a great podcast to do that because we get to do that here too. get to be critical of the profession because we love the profession and we want to see it improve and get better.

And these are the conversations that need to be happening at a broad scale. and so. , do you have experience with, companies, on, no matter where they're coming from, who do see those advantages and maybe just some examples of how they've been able to enable their business to do better by doing that, by engaging with you the process so that they can excel the market.

Stefana Parascho: Yeah, so I haven't really worked directly, uh, with, construction companies for example. But I've seen a lot of projects that, were more targeted at the direct application and from my colleagues, basically, and I, I wanna say in Switzerland, for example, the timber industry is, Very interested in the digitalization process for good reason.

and there is more and more companies that, um, they build up their own little robotic labs. and they might not use those robots [00:42:00] to build these unique geometries that have never been done before, 24 hours a day. But the machines are still used, so they find use for, more, standardized construction too.

But they have this possibility a project comes in that requires more complexity, they can immediately set it up and do it. And I think there is a strong interest there. and it's not that, unaffordable. Like these robots, sure, they're expensive for a private person, but they're not expensive in comparison to any sort of processing machinery.

Like they're way cheaper than a good CNC mill, for example. 

it's more the, the skills that are required to then program them to define what they do at settle up tho those obviously, they cost money time. So that's a little bit I think that's the bottleneck in, robotics more widely used in industry is either making, requiring less skills to do it [00:43:00] or, having more people, come out with those skills.

That can be, quickly employed. I, think what the feedback I've heard is that it's also really difficult for these companies to find the right people, is fair. We educate, uh, Certain amount of students, but they all, they also typically don't come out of architecture school as specialists in controlling robots.

and there is usually just a handful that, really might dive into it. So not a saturated market there, for sure. Somehow some of the, a lot, a lot of the, maybe the teaching that we do is more, here's an introduction and what are the potentials of it. And many are excited about what the tools can do in terms of oh, do, can now design differently and I can think and generate of different things.

but it's still quite a niche field, for somebody to choose that as a career path and be like, I will be the specialist who, really control these machines and build up a career on that. So it's not. quite there yet.[00:44:00] in that sense.

Evan Troxel: right. Well, it 

does seem like it is 

inevitable though.

Because specialties expertise doesn't have to be like full-blown expertise. It, it has to be 

this exposure to it because this 

kind of natural architectural mind, that student, it's, it is lifelong learning. They will 

continue to develop and take in new inputs that they get exposed to outside of, of school, but then apply what they learned in school to that and that that saturation will happen kind of organically over time.

I, there 

there is some exciting possibility there for sure. But how many students do you guys have going through your program that, that come out 

with what you would, you know, with a somewhat of an expertise in robotics?

Stefana Parascho: So that changes from school to school, from context to context.

Evan Troxel: Hmm.

Stefana Parascho: Um, and I wanna say it's, it's becoming more and more, um, let's say to put a number on it. Uh, it's a bit tricky, but maybe like, don't know, [00:45:00] 20 a year might have passed in like a general, in a generic master's program, architecture, master's program.

maybe it's 15, maybe it's 20. Depends a little bit on, on the year, the school and the capabilities. Like any P F L, uh, I'm one of. Maybe two professors who do anything with robotics and architecture. There's other schools where there's more going on. Um, but I do know there's a shift luckily, to introduce any sort of digital tools way sooner in education.

So there's, uh, there's schools that have, they, they already teach programming, any sort of digital design fabrication in a bachelor's level. We don't do that here yet much. There's a little bit going on, but not much. But other schools have really implemented this as like fundamental courses. Everybody learns grasshopper, everybody learns a bit of programming, and everybody gets exposed to it really, really [00:46:00] early.

Uh, which I think is a really nice shift.

Evan Troxel: Nice.

Stefana Parascho: so yeah, in the long run it's gonna be more and more and more that for sure. right now I do still feel like it's a bit of a niche, but

Evan Troxel: Yeah, 

for sure. I, the, I, I do see it though in many schools, have some form of. A robot in their program. 

I don't know the utility, the u you know how much it's getting used, but I, I had Madeline Gannon on the show and she talked about, you know, finding a robot in the basement of a lonely robot in a, in a dark basement and, right. And, and so sometimes that happens, right? The, the school gets the funding, they buy the machine because they have to spend the money to do it before their budget's gone. And then it just sits there because no one knows how to use it. No one knows what they want to do with it. But then there's somebody who comes along and is like, we're gonna start using this thing. and and I can imagine that, that that is how a lot of robotics gets introduced [00:47:00] into, into schools, whether they're architecture or engineering or whatever. But there are schools like yours that actually have a dedicated program and people go to school there to learn that specifically. So there's also that also contributes to the lack of saturation in the market. Right. That you're saying there's only 20 graduates per year you can imagine, is more schools uptake with this kind of a program that that's gonna change things as well? 

Stefana Parascho: yeah, yeah. I do think it's a, like in anything, no, economics, it's a, a balance between demand and, uh, offer, so to say, because we do on, on the other hand, sometimes hear the critique that we're not educating, we're not preparing students, uh, well enough for the reality of an architecture office. 

Um, and then it's like, yeah, we're not aiming for that because like we wanna change that reality.

But it's, it is only, yeah, it's always a bit a [00:48:00] question of how many, like, can, can they actually find a job with this and the job that they're happy with? Um, with, with the education we provide or is it, it's, yeah. The whole process moves hand in hand, right? it's, uh, 

Evan Troxel: it it is interesting to think about academia being behind in that regard 

m much of the time, which is exactly why the industry is taking so long to adapt 

and evolve, right? Because schools have traditionally trained people for the last version of practice, 

right? 

Which, and, and and it's interesting to hear you say like, that's not what we're aiming for.

We're aiming for the next version, right? 

So preparing students for what's coming, not where things have always been, is uh, I think a very refreshing view to hear, and I'm, I'm happy to hear it.

Stefana Parascho: Yeah. . Definitely.

Evan Troxel: So, I think one, oh, maybe last question here we can wrap up is what is possible now that wasn't even possible about like five years ago in [00:49:00] robotics and architecture with what you're seeing.

Stefana Parascho: I need to think about where things were five years ago. It's not so long ago.

Evan Troxel: It isn't, but, but 

covid happened and so it's also an eternity, right?

Stefana Parascho: Yeah, it's

Um, let's see what's possible though, wasn't possible. Well, I do think access to both the knowledge and even just the hardware has become a lot easier. Um, In terms of, I, I'm trying to think about where was I five years ago and I was at E.T.H. Towards the end of my PhD. Um, there were four robots in the hall and a handful of people who really understood the control and the software.

And now the same hall at ETH, I don't know how many robots it has cause I stopped counting them. [00:50:00] But, um, I wanna say 10 or something a lot more. There's a lot more going on. Um, machines of different sizes. Um, there's control software, open source control software that's been developed and that's reaching a point where, Um, yeah, people can use it a lot easier.

So the, the, the entry to actually using it, uh, is, is it's a lot, uh, more accessible. that I think would be the main, uh, the main change. And like you said, so many schools have a robot or several just sitting there. I think that wasn't the same maybe five years ago. Some had, um, but like 10 years ago, definitely not many schools had a robot in their basement.

And right now they all do. So it, there is, uh, an increase in the accessibility of the technology, which, uh, yeah, I think enables a lot more [00:51:00] to happen. There's also so many more people doing this type of work and research. Um, when I looked for a PhD position, I think. Around here, I was only considering two schools as an option, and right now you could have, there is, yeah, so many more that I could consider for my PhD.

So it's, it's definitely, uh, expanded quite quickly and, given us. Yeah. Given a lot more possibility to more people, to do more research, to build up on things. I do think it's also become more, on the one hand, more, there's more in-depth research happening, tackling really, really tricky technical issues, which maybe at the very beginning it was more about just make it work, , make it somehow do what you wanted to do.

And now because. There's a wider range of, of topics of people of [00:52:00] knowledge out there. You can't really go deep and solve something that's been nagging you for years and years, but you just never had the time of possibilities to do. And on the other hand, there's also a lot more thought about the applications.

There's also a lot more applied research connected to industry that's really going into industry, et cetera. So, um, that's all abstract things that I'm, uh, listing just to say that a lot has happened. Definitely. And it's in the impact is increasing, the range is increasing. Um, yeah. I, I think at, at, at the same time, if I can also, like I give a bit of a critical outlook, it's, it almost feels a bit saturated in the sense of what are we doing with this technology.

So at the beginning it was really exciting that we could build new forms and use. Different materials in different ways, and at this point it's like, yeah, we've kind of seen it all, or if we haven't, we know it's possible. It's just [00:53:00] somebody hasn't yet implemented it, but they will in the next five years.

So it is a bit of a changing point of what, it's maybe not enough to just prove that we can do something that is possible, but um, it's a point of really questioning what is this good for and what is the impact and where, where should we take the field so that it has a real impact beyond exploring new things that weren't possible before.

Evan Troxel: Yeah, that, that's,

really great to hear because I think that is typically kind of a critical missing piece of the 

puzzle, which is what is the long term?

Stefana Parascho: Yeah.

Evan Troxel: Trajectory that we're aiming for, not just 

the short term, like creating cool forms that were previously maybe unbillable, I mean obvi, all those are pieces of the puzzle, but like 

at the puzzle scale, at the big scale, what, where are things 

going and how [00:54:00] can we help steer get it to 

where the ultimate vision is.

That that's great to hear.

Stefana Parascho: Yep.

Evan Troxel: The, the idea though of for students to, I think it's gotta be so exciting to not only kind of, because of the tools that exist now that didn't exist. You know, let's just go back 20 years or maybe maybe 25, right? The tools that exist now that enable. The actual fabrication of the built environment through robotics to create very precise, that's probably a redundant way to say it, , it's just precise, uh, construction constructability of, of an idea is gotta be so satisfying for students to be able to participate in that from ideation to actual inception of that as a built thing that.

I mean, unless you took a construction class and you learned all the, and [00:55:00] people have wood shops and they've got welding shops and stuff like that, but to go beyond those, that typical kind of construction and create, even, even, you know, going back to the pavilions, it's abso, the thin shell pavilions, the carbon fiber pavilions, like the 3D printed pavilions.

There's so many amazing things that are possible now. To see that happen in school has just gotta be so incredible to actually write the script. You know, start with the sketchbook, write the scripting that and perform, you know, builds the geometry and then you kind of reverse engineer the paths for the robot to follow, to do the thing and then actually do it has gotta be so amazing to go through and such a, an incredible experience, a formative experience for architects of the next generation.

So, I mean, that, that to me sounds super, super exciting.

Stefana Parascho: Yep. I, I can subscribe to that. At least. I, I was really that generation [00:56:00] that got to see that maybe first or like between the first, one of the first generations to really see that happening. And I, I, I think it was mind blowing. I, I constantly wonder if it's the same for my students today or if, because they've kind of seen a lot more of it if it's not as exciting anymore.

But I do think it is, I mean, you can see them when they, even the smallest thing that we do in a. In a seminar here, you know, a two hour a week sort of course, when they actually get to see the robot build, do what they wanted it to do from the beginning. I, I, it's, I, I, I wanna believe, uh, I wanna say, I can still see it in their eyes that they're pretty, uh, blown away.

So. Oh, , I'll believe it's still happening.

Evan Troxel: Great.

Stefana Parascho: Yeah.

Evan Troxel: Well, Stefana, I think that we've covered a lot of bases today, and I, I appreciate, is there anything else that, that we missed that you wanted to [00:57:00] chat about or expose the audience to regarding what you're working on?

Stefana Parascho: Oh, there would be many things I could talk on, but I think we're good for today.

Evan Troxel: Great. 

I'll put links to everything that you're doing online so people who are interested can, can learn more and learn more about you. And I really appreciate you taking the time to have this conversation today.

Stefana Parascho: thank you. Thanks a lot for having me./