Martin Bielicki: Hello everyone, my name is Martin. I'm the founder of Bench, and today I'm joined by Paul in our Blueprint podcast. The goal of Blueprint is to disseminate and democratize AI knowledge in engineering. and today I'm joined by Paul, a teaching fellow in industrial design at Loftboro University and previous previously a design engineer himself. He's got a lot of thoughts and experience with using AI specifically for CAD and product design. So we're gonna be speaking around those topics today. Paul, I'd love you to introduce yourself as well.
Paul Russell: Hi, yes. So I'm currently the programme leader for the masters in integrated industrial design at Loughborough. I teach heavily into the undergraduates as well product design and technology and industrial design.
Martin Bielicki: Awesome. Paul, let's jump right into it. So you've taught CAD for for years now and you've been in Loughborough where Chat GPT first landed and AI started really picking up. What have you seen change since then?
Paul Russell: well I would say even before even before, say chat GBT, I mean I was in the image space before the tech space. And I remember even like January twenty twenty one, you know, there's there was that demo from it would have been the first version of DALI, but even before that there was things like disco diffusion. There there were other things that just weren't doing anything, but but the second of that there was a demonstration where they were doing like, you know, chairs in the style of like an avocado. And that for me was pivotal in that, you know, you knew that was only early days and there were things there like the abstraction and the semiotics, the ability to embody it and to visually communicate it. And you think, well, okay, if you put that in front of a student or a designer, you know, they wouldn't bring that many concepts with that many much variety, as quickly. E even including the fact that, you know, things were touch and go, then y maybe one in five images was good. But the point being, you know, you could generate a hundred images and twenty of those would then be good, right? So that was really eye-opening and that moved fast. You know, I was in the betas for mid-journey in early 2021. And then things like stable diffusion dropped around the same time as Chat GPT. that was just really interesting. again, that kind of okay, maybe one in five images is good, but there's enough in there that it's valuable to kind of spend the time running this. and on the on the language side, you know, it was really interesting, even early days, GPT three, three point five, you know, ask asking it to kind of write, you know, five hundred words about some area of design practice. And there was enough in there, there was enough eloquence, there was enough understanding. And the and the the way that you looked at it, it's people people would look at outputs in the early days and compare it to their pit themselves and their peers. it's not that good. But you've got to understand there's eight billion people in this world, right? Probably seven billion of them are nothing about design. more than seven billion I'd say. it's it's that kind of narrowing of of of the kind of knowledge and understanding in that things are now kind of well, I don't want to get into semi s semantic debate about the word understanding, right? But the ability to communicate things and to perform kind of economically valuable work, it it is happening really, really fast. and we've got to keep on top of that.
Martin Bielicki: So you've seen the writing on the wall very early, right? Because I think
Paul Russell: Yeah.
Martin Bielicki: even now, I think engineers are we are very I guess skeptical species, you can say.
Paul Russell: Mm-hmm.
Martin Bielicki: And I think a lot of engineers, whether software or even more so hardware,
Paul Russell: Mm-hmm.
Martin Bielicki: still don't believe that AI can do their work. You know, whether it can do it fully, it's a question
Paul Russell: Mm-hmm. Absolutely. Mm-hmm.
Martin Bielicki: for another discussion. But why do you think you you've been so early to me. So why do you think you Like why did the skeptic in you not not kind of start kind of l sounding alarms and why did you believe so early that this will change fundamentally how engineers work?
Paul Russell: Yeah. I think I think in if you look at trajectories, you look for back from three ch chat GPT three back to two and you see that leap and you think, okay, even if the pace continued at this rate, right, and it wasn't exponential, it may maybe it is, I'm not that's not another another discussion.
Martin Bielicki: Mm-hmm.
Paul Russell: It it would only be ten, twenty years down the line that things would develop and and I mean I I I have been into kind of human intelligence since was very young. so I I and artificial intelligence as well. And and you can see If you can if you can codify things and see patterns, then you can kind of do it with a computer effectively. And a lot of what we do can be can be codified and you can you can see patterns in it, you can codify it, you can replicate it. and I think that's true of a lot of things that people don't want to accept are going to be
Martin Bielicki: Mm-hmm.
Paul Russell: true. And a lot of that is more human centered kind of things. But you know, talking about being an engineer, you know, I I straddle both sides. You know, I have an undergraduate in design, I have a masters from an engineering school. I teach kind of across I've worked across both, teach across both and You know, I I I I'm kind of foot in both fields and and I think you know, there's a lot of skepticism in design as well, but it's a different there are different kinds of skepticisms as as design engineers and engineers. and the reasons why, you know, people are adopting it and and resistant to it, kind of some are shared and some are different, but it's good to have that full spectrum and I think that that helps me as well to see like where it's gonna land and where it's gonna fit in because I I'm teaching across the board, you know, in terms of front end research. ideation, development, you know, the CAD rendering, technical drawings, you know, design for manufacture. And and I can see opportunities throughout the you know, across the spectrum for AI to be incorporated.
Martin Bielicki: Let's move to to the to actual engineering from from I guess from images, I guess. So how
Paul Russell: Mm-hmm.
Martin Bielicki: have you seen that shift play out of of AI starting to enter CAD and technical drawings, all the other parts of
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: the design process? And how have you seen it impact not only your students, but then the industry? Because I'm I'm sure you have ties to the industry at the university as
Paul Russell: Mm. Yeah. Yeah.
Martin Bielicki: well.
Paul Russell: It's very well, it's a very mixed bag at the moment, of course, because we have we have mesh models, like AI mesh models, and they are incredible and and I suppose that's slightly one of the criticisms I would have of of some design education is that if you can produce a fancy few renders at the end, right, and and a a G A, a basic G A, you know, you you you kind of jobs done, it doesn't go necessarily to the level of design for manufacture costing and you know, being able to do be produced at scale. So I think, you know, if you separate the two and then you look at you could you can you can do that today, right? You can make an image, make a mesh model, render it from a couple of angles, right? And you don't need that CAD model if you see design as being like, okay, this is like, you know, the the the the old notion of handoff to engineers, right? And in many ways that's not the case so much in the UK anymore. I think we are one of the most forefront kind of countries in terms of having that like link between design engineer engineering. But other countries, you know, there's still the concept of designer handing off to engineer and and then the engineer would have to model it properly, et cetera. but I think in terms of engineering, it's I think we're seeing early days of lots of ideas, lots of competition, lots of noise, lots of mess. And I think that when you look at what's out there, and that there are a couple of big companies, you know Bench is one, Leo is another. what's the other company called Backflip or something like that?
Martin Bielicki: Backflip, yeah. Mm-hmm.
Paul Russell: Backflip, right? Yeah. they're all looking at it from slightly different ways. I think Backflip's got an interesting approach. You know, they're looking at converting mesh models into parametric models. And that that's gonna have that will definitely have some benefit in that area. But you look at other companies that are trying to do CAD straight away. Some are doing it inside APIs, some are using existing kernels, some are some people are trying to write kernels. And there's loads of things happening. And yeah, it's difficult to say where things are going, but I do I am conscious that there are three big players, right? There's PTC, there's Autodesk, and there's Assault. And they particularly Audest, let's say, have have a a habit of buying things up when it kind of threatens them. And then they kind of you you arguably maybe they stunt innovation sometimes, or they integrate it fairly well, like they did f with T spines, for example.
Martin Bielicki: Mm-hmm.
Paul Russell: showcase and fusion and things. But other things they kind of buy it and and let it go. and DeSalt have, you know, they are not I don't know they have a hesitance to buy things because I don't know they feel like they can just do it better. And then PTC, of course, have just bought OnShape, but they're also par in parallel at the moment, trying to write like AI for OnShape and AI for CREO. They're completely different kernels, you know, they're like doubling up their own work. And I think there's a lot of mess. And I think that what what you see you're gonna find there are limitations not only in terms of the level of access that companies will give to the API, but I think just in terms of even kernels, you know. I mean if you think about CAD, like I I've been thinking about this recently. You think about like SolidWorks, right? So it uses Parasolid, but so does Shaper 3D, so does Plasticity, so does et cetera. And and if you think about like the kernel, the kernel is like a model, right? And then on top of that, you have the API. And the API is like a harness, right? And when you when you think about things that way, you can see a lot of the problems aren't necessarily with the kernel, they are with the harness, right? And you could but you're gonna have limitations where the haun the ha the model and the harness aren't being made by the same company. That level of integration will be limited. and I think the real potential then is for players to come in. And build something that is a modeler from scratch. You know, it can model, it can do assemblies, it can do simulation, PLM, but the whole kind of works, right?
Martin Bielicki: I think that's that's the golden really outcome. I think the difficulty
Paul Russell: Yeah. Mm-hmm.
Martin Bielicki: is the first thing that corporates tell us currently when we speak to them is I don't wanna switch away from CATIA I don't wanna switch away from Siemens NX It's the first sentence we hear. So that's
Paul Russell: Yeah, yeah. Mm-hmm. Mm-hmm.
Martin Bielicki: the difficulty, right? They're obviously imperfections and these
Paul Russell: Yes.
Martin Bielicki: softwares are obviously not optimized for agents to use. They're not AI native software
Paul Russell: No, no. No, no.
Martin Bielicki: in kind of whatever definition we're gonna use.
Paul Russell: Mm-hmm.
Martin Bielicki: But also we have to meet the industry where where they are. And kind of
Paul Russell: Yes.
Martin Bielicki: the the cost of retraining a whole workforce of BMW to, you know,
Paul Russell: Mm-hmm.
Martin Bielicki: a novel AI CAD, as much
Paul Russell: Mm-hmm.
Martin Bielicki: as it would be probably greater at some point. I do believe it happens. When does it happen? I think it's
Paul Russell: Mm-hmm. Right.
Martin Bielicki: not for another couple of years at least. And I think that benefit would have to be so clear for them to to actually undertake
Paul Russell: Yeah. Yes, yeah, yeah.
Martin Bielicki: that that motion.
Paul Russell: Yeah. And then on the on the other side you've got I mean, I have seen more new CAD programs and attempts to build kernels in the last two months than I've seen in the last ten years. That is happening really quick. There's a lot of companies out there making parametric modellers, basic ones, you know. But so some are using O C C T but some are building like kernels. some are making kind of brand new surface modellers from scratch, you know, they replicate things like Rhinos or alias even. and I think you'll see a lot in that area as well. And that's another interesting part because you know we talk about mesh, we talk about parametric. But there is also the kind of okay, are we talking about you know simple you know feature tree parametrically modeling parts, or we or we're talking about surfacing because there is a difference there as well. And think of course I think we're we're we're probably yeah we're definitely less than a year away from you know the standard kind of e exercises you get in university like here's this fully dimensioned three view drawing model it in CAD, right? One click to get that in a in a step file. That's well far less than a year away. And when we
Martin Bielicki: Yeah.
Paul Russell: can do that, you know, how does that impact how we teach CAD? If a student knows I can just one click it, well where's the motivation there to kind of learn it? But they need to learn it to have the basis to do more advanced things. And that's going to be an issue
Martin Bielicki: Because
Paul Russell: we face.
Martin Bielicki: yeah, I guess that's my next question. How I think that's a broader question about academics. Like, what do we teach now that AI is here? But of course we we
Paul Russell: Yes. Yeah.
Martin Bielicki: should focus on on design engineers and and cat in general. How do you approach this? So how do you approach teaching your students to enter this new paradigm well
Paul Russell: Mm-hmm.
Martin Bielicki: with the necessary skill set of the kind of bedrock
Paul Russell: Yeah.
Martin Bielicki: skills you need to be an engineer? While also not them them not feeling like they're they're doing manual work and they're kind of twenty years behind what's what where the industry is.
Paul Russell: it's not threatening CAD just yet, but I think I was I think there's still possibility by October, for example, that the one that they have inside Fusion is up, running and pretty functional in terms of you can use natural natural language, you know, build this block, fill it these edges, and it will understand, you know, if I have like a block and I say fill it the vertical edges, it will have enough semantic understanding to pick up what what that means. And I I think I would say that probably by October next year we'll have a lot of students coming in with fusion experience that have experience with the AI as well. And they probably will see quite mm that there there definitely will be limitations in in how complex it can be. But yeah, I mean the the the difficulty there is there's another factor in education in that there's quite a lot of students. I mean we have you know, we have say you've got five hundred students, you know, you'll have hundreds that are fantastic, great, brilliant. Th there are students that just wanna do a degree, right? They just wanna get through the degree. And I think those students will be able to do that with AI, even with CAD by next year, next October intake. And and how do we manage that? Because then what are we you know, are we delivering an education or are we allowing them to kind of jump through the hoops to kind of do enough work to kind of get a degree? And we nobody wants that. You know, it's not good for the students, not good for academia. It's not good for, you know, people that that, you know, are arguing that there's too many graduates already anyway. And you know, there's there's a lot of factors involved there. but how how do we then Yeah, I mean I I think it the same with a lot of other stuff that I'm doing that, you I do a lot of I mean, I've taught for years key shot and blender rendering, right? Or really high end stuff in Blender. and we s I you know, Gemini Nana Banana came out and then Veo and you're like, my god, like this is insane. The the the speed and the accuracy and the context and the realism, even if you hash it at five, ten, fifteen times, right? It takes dozens of hours to make work in Blender that that's not even of that level of quality. And that's something I've had to face already, right? So as an example. and I I had to kind of bite the bullet, you know, I took years to learn Blender. I enjoy Blender. Actually, and that's another subject I'm sure we'll come on to, you know. What what about the enjoyment, right? Like I enjoy using CAD. but I ha I I'm not gonna hold back and say we must use KeyShot to make subpar renders because I I don't know, it just doesn't make sense. You look at the history of of what we do, you look think about drafting, right? I I my first design engineering job. There was a draw full, like a planar chest of like A1, like hand drawn, hand lettered written out drawings. Nobody does that anymore. We can't nobody can really do that anymore. and we don't consider it time effective, effective use of time. It's not cost effective. It's not, you know, it's not robust enough in that, you someone spills coffee on it, you're in trouble. so things things do change and and we have to evolve and adapt and and I'm trying to kind of keep on top of that. And actually not only that, but be slightly ahead of the curve because You have to understand as well that most of our students do a placement. So a student entering this year will leave into the workplace in 2030 in four years' time. I am not training a student for work in twenty twenty six. I am training a student for work in twenty thirty. And that is a really critical thing that that most educators haven't quite gotten to yet.
Martin Bielicki: And that is also so hard that you to try to predict that future, right?
Paul Russell: Absolutely. You yeah, and exactly and so and sometimes it is i it's a bit of a gamble, some of it will be right, some of it might not be exactly right.
Martin Bielicki: Yeah.
Paul Russell: but you have to do that. You can't put people out into industry that, you know, are are designed you know, prepared for industry as it was when they entered.
Martin Bielicki: On that topic, you mentioned that AI can make good designers better, but it it won't do much for mediocre or bad designers.
Paul Russell: Yeah.
Martin Bielicki: Why do you think that and how do you think that that plays out over the next couple of years?
Paul Russell: Yeah, I mean I I wouldn't say only AI. I mean perhaps it's like technology like like having access to like renderers, like being good at rendering, is better than for example, you know, old days of Photoshop rendering and before that marker rendering, before that chalk and pastels, right? So technology helps people that that kind of master them to get even better. And and the same with CAD, right? What's the alternative to CAD? You know, your hand drawing, technical drawings, right? you know, the speed, the efficiency, the kind of ability to I mean s some things, you know, that that are very difficult to draw, can easier to CAD model sweeps, lofts and things, right? so th not just AI, but CAD helps getting good at CAD helps good designers get even better, right? And if you just kind of learn the basics and only pick up the basics, you know, we w you see students that graduate only being able to do extrudes and fill it, right? That that happens, right? And some of them, to be fair, get to relatively complex forms. Granted they're only using like extrude boss, extrude cut and fill it. But they get through it, right? They get jobs done. But being able to learn, for example, you know, more advanced tools, both in solid and surfacing, does enhance their work. Same with AI, right? But if you engage with it, I mean I've been teaching AI for many years now since 2022. And I suppose it it's no different than anything else, but but you you can learn it passively at a surface level. You can kind of tick the boxes or you can really engage with it. You know, a student could and an interesting one, right? So it ca it takes for example, some of the modules this year, right, using flow, it's three hundred percent free, so there's no cost to it. You open a web page, you put a text prompt in, you get four images back, right? It takes twenty seconds. You could do that, you could spend five minutes, right? And you could have hundreds of images. And even better than that, you know, like I was saying, four years ago, one out of five images would be roughly usable, right? Today, you can make one image with five design concepts on it. Of which a handful are usable. And you're making four images at a time, you look at the scale of that, right? And that'll only increase the the capacity to do things like that, to get ideation, inspiration, right? but you know, one student might do a hundred images, one student might do f five and feel like they've ticked the box, right? And then it's that level of engagement. But it the more that you use it, the more that you learn from it, the more that you learn to work with it, the you the more that you learn how to use it, how to engage with it, how to share agency, how to co-design effectively and and that's gonna happen a lot in CAD too, right? So if you think about CAD, like before it existed, a lot of the kind of idea was that it was computer aided design, right? The computer would aid you to design. But we've used it for, you know, four years as a drafting tool, right? It's a digital drafting tool. It's not it doesn't aid you in any any way. There's a plug in for SOLIDWORKS like DFM Advisor and and it's it's very primitive compared to what we could already do, right? And Autodesk Fusion has one and they're working quite heavily on it. But I think that is where the future goes as well, right? You might model a part and the the AI might be constantly monitoring that part and saying, No, actually that wall thickness needs to be thicker, thinner, you know, the ribs need to be thicker, thinner, you need a boss here, right? And people will have to learn to share agency with, you know, the computer or, you know, the the the AI inside the CAD, just like they will, you know, things like research, you know, front end research, secondary issues that is you know ideation inspir you know getting inspiration you know all kinds of things but it will happen in CAD too and tech like technical
Martin Bielicki: I think
Paul Russell: drawings as well right I mean that's going to be an even easier one right you've got drawing and it's gonna say well you're missing the dimension here or that doesn't have the right tolerance or you know th these things will happen and people will have to get comfortable with that.
Martin Bielicki: What do we generally believe is that, kind of on the topic of enjoyment as well, but kind of I think even more so impact, is that broadly engineering is going where software went. Where essentially now engineers and software
Paul Russell: Yeah, yeah.
Martin Bielicki: engineers essentially manage swarms of agents and essentially ride the system that directs the agents.
Paul Russell: Mm-hmm. Yes, yeah. Mm-hmm. Mm-hmm.
Martin Bielicki: And that essentially what we've seen kind of and kind of how our our software engineers
Paul Russell: Yeah, yeah.
Martin Bielicki: code, the enjoyment is very much there. It's just the pace.
Paul Russell: Mm-hmm.
Martin Bielicki: has picked up so much.
Paul Russell: Yeah, yeah.
Martin Bielicki: And it's it's basically the limitation seems to be the the cognitive abilities of the humans. Essentially how many
Paul Russell: Mm-hmm.
Martin Bielicki: agents can you manage at the same time? Because technically you do infinity. But then your your context
Paul Russell: Yeah. I think it's gonna be a bit slower yeah, I think the pace will be a bit slower for engineering. But I I do I have followed, you know, heavily like coding and programming, you know, how how people are accepting it, adopting it, the resistance and you know, there's resistance, resistance, resistance until the the the system's got to such a capable level and lots of people now like, Okay, it's changed, right? There's nothing I can do about it, right? It's yeah, it is what it is.
Martin Bielicki: There's no question now. There's no question. Like two years ago we had we had skeptics and now I I've I very rarely get any s skeptics on on AI and software.
Paul Russell: I honestly I I've been very skeptical of AI CAD myself, but I've seen enough now to know to I can see where things are going. I can see the patterns, I can see the trajectory, I can see, you know, how you can emulate lots of things. So it will it will happen. but Catia
Martin Bielicki: Wa
Paul Russell: Catia and Siemens, sorry, are are already quite they're a bit down that line of agents, right? They already have abilities to kind of request and and constrain things and get results in CAD, right?
Martin Bielicki: I think they do. Yeah. Catia's Catia, for example, Dassault are working on on their Leo agent, for example. They've got a couple, I think three agents. I forgot the other two's names, but but Leo is the one doing doing CAD.
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: for sure it's getting better. I don't know the exact capabilities, but yeah, it's getting better and better. why do you think the progress is gonna be slower in in engineering? Do you think it's more so a technological block or a cultural block?
Paul Russell: Yeah. I mean it's it's another dimension, isn't it? I mean I suppose, you know, even with programming to an extent you can work in in three dimensions. You could you can design games, right? That that are in three dimensions, applications that display graphics in three dimensions. I don't know, I think that there is more to it in terms of materiality and I think, you know, simulating a piece of software is relatively simple and I think I mean it's it's more data, I suppose, in physical side, you know, everything must have typically kind of some level of i it required physical strength, you know, weight limitations, heat transfer, you know, there's gonna be a lot and and costing as well, right? So I mean a lot of, you know, what I teach when I teach design for manufacturers, okay, you've got like you've got like cost, you've got strength, you've got aesthetics, you've got like other human factors, like is it cold to the touch, is it comfortable to hold, you know. And the like the intertwining web between all of these things is so complex. I do think it's a level of complexity greater than than programming. And I think what you'll see, I mean I think you'll see things like if you look in the large language model space, you've had people that trained initially like coding models, right? It did coding, you did like ones that did like natural language literature poems, whatever. You had ones of people that would do it for certain different things. I'm not gonna use the term of general intelligence here, but you know, you see things over time like come together. I think you will see things in isolation in engineering physical design. And I think it'll be a little bit of time until they start pulling together to create kind of models that are generally capable in in engineering.
Martin Bielicki: Mm-hmm. Let's let's pull that string a bit. you've got a a course about the ethics, the ethics and aesthetics of generative AI in design practice. Could you tell us more about that? Kind of how how ethics and aesthetics are intertwined intertwined with generative
Paul Russell: Yeah.
Martin Bielicki: AI.
Paul Russell: Yeah, I mean so I think I think what a lot of it was is you have a lot of people that say, these these images are great. you know, the argument is are they good and are they not good? And then you have like people that are arguing, you know, are are they is it ethical or responsible to use AI, training data, etcetera? is it unethical? And I think those conversations we've maybe I had in kind of isolation, you know, they're really you know, to get everyone to kind of some level of like understanding of everyone's position. You know, I I'm in the module I'm happy for people to say and and and more so in graphic design, because a lot of students take it from graphic design as well. You know, they say, I I don't think the results are that good. I th I think I'm better I would generally disagree, but that's a different subject, right? But they're allowed to have that opinion, right? And and that's fine because the point is that they leave the module with an informed opinion, not that they leave in harmony, right? Agreeing with one another or with me. so I think, you know, you look you know, some people can look at the images and they could say, Well, they are beautiful, but actually is it you know, ethical and they could say, Well, you know, it's unethical. Sorry, sorry, the images are terrible and and but it, you know, it is ethical and and I think there's a kind of crossover there where everyone's having kind of two arguments across one another and I think they are more linked than you realise. I mean ethics is sorry, aesthetics is a form of ethics and ethics is a form of aesthetics. They're not two different things, right? So my my real plan with that was to kind of get people generating content, evaluating you know only not only the quality of the content, but using a variety of models, comparing the models to one another. and then also I teach environmental kind of impacts of AI, both on the hardware side, you know, that we need more graphics cards. We have to force people to mine the Congo and and other countries, and that's not great. you look at the socioeconomic side, okay, so okay, it can emulate work and maybe replicate and and kind of replace work eventually. Where does that leave us? What do we do? Do we reskill? And and also, you know, does it reshift the location where people work, right? So maybe there's lots of people working in the UK. You look at say data centers, they were outsourced, you know, Malaysia, India, et cetera, et cetera. You know, if someone if their only duty is to watch agents and kind of manage agents, do we need to be paying them, you know, high salaries? you know, all all these kind of questions that that that need to be had. and also like I it's an opportunity as well to bring in intellectual property, copyright, patents, design rights and things, 'cause that's not covered too well generally in education.
Martin Bielicki: What we're seeing in many areas of of work done that is a huge bifurcation of the top ten percent getting paid more than ever and are more valuable than ever, and
Paul Russell: Mm, mm, mm, mm.
Martin Bielicki: the bottom ninety are trending downwards.
Paul Russell: Yeah, yeah.
Martin Bielicki: my guess would be the same happens in design and engineering, as
Paul Russell: Potentially.
Martin Bielicki: as you can scale up the impact of any and like the best individuals somewhat in infinitely.
Paul Russell: Yeah, yeah. Potentially yeah, I th I think it's two conversations there. Th the one the one thing that you that people really need to be aware of is a lot of the people that are taking a hundred million pound a year jobs to train AI models are doing so and the salaries are so high because they know there won't be any jobs in programming in a few years, right? As soon as we have self kind of training models, they know there won't be jobs, so they have to take those salaries now to kind of be future proof to to to have money to kind of get through, right? So we we know that's happening and and so it might it might look like okay, some people are getting paid loads of money, but I think they know as better than anybody that the money is high because they are training the models to the highest levels to replicate the work that even they do. The flip side is you know, a scenario where we have global power shortages, GPU shortages, they'll put another world war which isn't off the table at at all. And you know, we lose these systems or we lose the the access or all kinds of things or or or politics, geopolitics that prevents kind of access to models. you know, I think if we do get to a situation where models can do everything and nobody thinks that anything's worth learning and and training on, yeah, we are kind of in trouble when if something happened that prevented us from kind of having access or using models.
Martin Bielicki: Yeah, and I I think that's where the the whole topic of sovereign AI owned by continents or nations or groups like the European Union, that's where this becomes becomes interesting.
Paul Russell: Mm-hmm. Yeah, yeah.
Martin Bielicki: Let's let's put our creative hats on for a second and let's try to imagine what does the job of a design engineer look like in five years? Well because obviously
Paul Russell: Five years.
Martin Bielicki: as you as you said, you you're just training your your students, Carney, for a century twenty thirty.
Paul Russell: Mm-hmm, mm-hmm. Mm-hmm.
Martin Bielicki: what do you think that looks like? I know you have a an interesting post on LinkedIn about how a design agency of the future can co could ta theoretically alterate more and more of its work, trending to like nearly a hundred percent in the coming years. How do you think that that plays out? Kind of what what are you telling your students and how are you advising actual design engineers that are currently in the market to to prepare?
Paul Russell: Yeah, I mean one one of the things I did, you know, a few years ago is I bought Blender and Rhino into the courses because I knew from the get go that e everything that was gonna happen with AI was gonna be in those programs first for a couple of reasons, right? Blender's hundred percent open source, open kernel. but technically Rhino is an open kernel as well. So Rhino the Rhino kernel is open source, so a lot of people write for that and and tinker with it and play with it. It's affordable. So a lot of you know the stuff I've done in recent years, you chat GPD, write some code, write some grasshopper dis definitions, do all kinds of things in Blender, right? And that's I'm not I'm not even touching on MCP or any of that kind of level of of automation yet. so one of the things I did was to say, look, you know, you don't don't understand CAD as parametric modeling, don't understand CAD as SOLIDWORKS, you know, taking things away from rote learning or skill learning or tool learning to higher level understanding of geometry, of computation, you know, things that you know that if you're gonna be mod watching kind of agents and things that you're gonna learn need different kinds of knowledge and understanding. so I think that's part of what I d I've done there. Now design engineering, I mean the things like technical drawings, right? I mean I as probably with all people, you graduate from product design, so my background is like product design, and you get sold this kind of dreams, you'd study and like you get like freedom, you get crazy projects, you get like design freedom. But the real world's not like that. and and you go into industry and you're like you expect to be sitting down all day designing CAD modeling and all these kind of things. And you f you realise, okay, I'm spending as much time an Outlook as I am in SolidWorks, right? But then there's a big chunk of time where you like engineering drawings and and even like, you know, CAM programming, CNC programming. Like I used to do a lot of laser and punch and things. And it's like they the the job's not that sexy, right? Some of it a lot of it is just not. But I think you will see yeah, I think this is the kind of worrying part, is you'll see a lot of that drudge work be automated and the drudge work is more easy to automate. And I think what you'll then see I mean there's two ways you can see it, right? Everyone just reskills and gains these new skills and suddenly the like the world needs more stuff. And I don't think that is going to be the case. and the flip side of course is that, you know, we need less people. You have five engineers in an office, right? You take The 30% of their job away at the storage work, right? You 30% of rest people. You drop down to three people, right? For example, I'm not saying this is going happen that the values are just like guesstimates, right? But that that's very likely to happen in many fields. So the purpose of what I'm doing as well is to say, look, you guys are getting those skills more so than any other degree I know. You are more prepared, more kind of informed. You have not only like skills and knowledge to like use tools and and do things but also, you know, I'm I'm doing that ethics and aesthetics. We're looking at philosophy, responsibility, you know, maintaining kind of design rights, human kind of touch, the whole kind of works really. but if you put me on the spot and said what does the future look like, I would say we are gonna have less jobs won't disappear, but if you take s a certain portion of people's job away and you take that away from everyone, you need less people to do jobs and I think that does seem to be where things will go.
Martin Bielicki: Yeah, I'm I'm curious how how it's gonna play out in terms of the market expanding. 'Cause just like I I kind of the the software market I make a lot of software parallels here, but the software market has expanded a lot after the I just because we now it's
Paul Russell: Mm, mm.
Martin Bielicki: now feasible to build so much software and people
Paul Russell: Yeah.
Martin Bielicki: can can now build one off apps here and there on their own because they can now you know vibe code, if you will, their own apps. I'm curious if if that that same will happen in engineering, where essentially if making stuff is faster and easier,
Paul Russell: yeah. Hmm.
Martin Bielicki: maybe it's actually now we produce more. it's of course a bit different. There's a limitation on manufacturing, producing. That's harder to access for a normal person. But yeah, that's an angle we also are looking at.
Paul Russell: I I would agree that there's gonna be definitely a democratisation of people that can go from an idea to a physical product. I don't think the world needs more stuff. I think there will definitely be a period where more stuff happens. And I think you know, you look at things like Terrible Places, Alibaba, Timu, that kind of route of kind of getting stuff out there. I think a lot of that'll be AI generated in in time 'cause a lot of it's kind of cutesy stuff, it's kind of, you know, bit of graphics on the world you know. Stuff like that, teddies and And all kinds of things, lots of that work will go to AI. on the high end engineering stuff, I mean I mean it's a sad reality that we live in, but right now we need more military equipment. Europe needs more, America needs more. you know, we we need that and and I think that there could be a lot more work in those areas in the in the next few years. And I think that's the kind of terrible human nature that that will be sustainable kind of careers to be in. But do we need More I mean if there were more planes out there would flying to America be cheaper? I don't know, I don't we need do we need more planes to we definitely need better planes, but yeah. Stuff like that.
Martin Bielicki: It's it's a tough question, right? It's a tough question. I think we we spoke to a lot of automotive companies recently and they said, for example, that the pressure they're getting is that China's producing a new car every year, like as if it's a new phone. And then I have
Paul Russell: Mm hmm. Mm-hmm. Mm-hmm. Mm-hmm. Absolutely, yeah.
Martin Bielicki: to pick up their design iteration, their time to market, just to really service that demand and be able to sell those OEMs without have to deliver a car every year.
Paul Russell: Yes. Yep. Yep.
Martin Bielicki: That's an example where I think we'll get more stuff. Whether we need it, probably a different topic.
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: but yeah, but even beyond military, we definitely have heard from across the industry that broadly in engineering there's more so a shortage than an oversupply of engineers. And and now especially with the cohort of
Paul Russell: Yeah, I mean that's that's a really interesting one. I mean for years in the UK you see it all the time not not enough people to fill engineering roles and you see things like you know engineer applied to loads of jobs, never got a kind of job in that. I don't know, it's a weird one. I mean so many of the jobs out there, like you go on like indeed or something, right? Most of the jobs are fake because they're recruiters and they're kind of fake jobs to get people's details on a system to spam them with jobs if they do come up. so is it really hard to determine real things like that. But the the reality is that you look at, for example, fantastic great universities like Loughborough, right? And say, you know, people who actually go into design engineering jobs directly, not like loosely related. It's probably about thirty percent, you know, you know the thirty percent that they go into, you know, loosely related roles might just be a CAD rendering or whatnot, right? I mean even like the engineering school f I mean you and I mean the statistics aren't really great, but the best that we have, you know, it might be fifty percent of say the engineers that get a job, right? But these are top top class kind of graduates from a great course, great university. So I don't know the notion that there aren't the jobs out there, I don't I don't sorry, there aren't enough engineers out there to fill the roles. I don't really agree with I know that that that the news says that and the statistics say that in the media and whatever, but I don't know if it's true. I don't know if it's true. Yeah, but but the the weird thing
Martin Bielicki: We hear it from from the industry though. I mm maybe there's something around from
Paul Russell: is, yeah, that they say that, but then they they kind of a lot of them have they refuse to deploy certain people and like For example, like an injection molding job right club and they like, I I'm only employing people who have got injection molding experience and that kind of mentality and then they don't recruit anyone,
Martin Bielicki: Mm-hmm.
Paul Russell: they're like, there's nobody out there. I mean that that is just for me is like absurd, right? same with machinists, right? Yeah.
Martin Bielicki: Maybe yeah, maybe there's a mismatch. Yeah, maybe there's a mismatch between what they're looking for, what they think they're looking for, and what's out there in market.
Paul Russell: Yeah, I mean no serious sometimes it's about it's like they don't want to take any responsibility or onus on for like training or development. They want people to come in the door, be able to able to do the job and I think that's an issue.
Martin Bielicki: Ready, yeah. Yeah, the I I can definitely see that. I think that there's probably something something something around there. Awesome. Paul, I've got one last question to close this off. if a seventeen year old came to you now and said they wanted to be a designer,
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: knowing all that you know, what would you advise them to do? Do you think they still need university? Do you think they can self teach things? Do you think they need to learn cat bottling from scratch on everything? it's a very broad question, but I'm curious to hear your take.
Paul Russell: Yeah. I suppose like my answer to that question would also depend on like how fast can universities evolve and adapt, right? I if the universities could adapt and evolve quick enough, definitely like university brings way more than the skills, right? And the knowledge, it brings the experience of like working in groups before you have to deal with like you know, you have experience working with people in groups and one might be not pulling their way or annoying and that's like real work, right? That's like that's like the real world and it prepares you for that. And there's so much more, of course, you know, like a degree. has a like an academic element to it, you know, you learn theory and philosophy and things like you learn to like a higher level than than you would just gain like the skills and knowledge. But I mean if I was seventeen today, w what would I do? I would like I mean I think it's I suppose the answer's true regardless, but learn materiality, just look around you, plu like use materials like wood, metal, you know, polymers, like look at what things are made of. Look, you know, gain Like because they're kind of there, yeah, I think in some ways like so my research is mostly on design evaluation now using AI to like enhance and augment like concept selection evaluation. And I think that's the kind of bottleneck right now, but I think that bottleneck will go. I think the bottleneck will be like can I get it made and manufactured? And I think the the design for manufacture knowledge and the material selection and and those and the costing kind of end of things are gonna be like the last bastions, you know what I mean? Like those last kind of real valuable skills. I mean AI will definitely get there eventually, but I think I think yeah, if you look at what's gonna be valuable, economically valuable for employers, the longest and the hardest for AI to kind of overcome, it is like the f feel, the physicality of the world. and I think that's what designers really need to be on top of now.
Martin Bielicki: Quick bonus question here. Do you think universities will be able to adopt in time? Because from my understanding, universities take three years to adopt a new new course or a new curriculum. And with the pace of AI change, three years is a whole era, right? It's a completely different paradigm.
Paul Russell: Yeah. But yeah, I don't know. I think if you look at higher education, it's like a slow moving kind of system and bureaucracy and things, you know, that's just d nature of it. And and it's the same thing like big automotive companies, right? Now I've I've heard a lot about, you know, companies that want to adopt Blender for modeling and rendering, right? And the the amount of kind of hubes they have to jump through for for what seems like a basic thing, it's free and open source, right? What's the problem, right? There's there's no cost involved, we don't have to cost it and all these kind of things, but they still have massive like systems and layers and and years of kind of getting things in. I think AI will be the same for a lot of companies, right? The b the big automotive companies. They'll they'll be very resistant. They'll have to do, you know, internal validation of security and all things. Can I can universities move quick enough? I don't know. I kinda I can say I don't know. I wouldn't say definitely I wouldn't say definitely not. But I think and I hope that AI does you know like the the kind of course kind of reevaluation, redesign, right? That that is definitely, yeah, you're right, like three, four years, years ago, I think that has changed. I think that things are more I think things are monitored more regularly now and and kind of redesigned, redeveloped, you know, pushed and pulled. I think if you look at the history of design education, things have always been that way. More things come in, some things kind of have to go out. But I think we have a problem in higher education now, in design education, industrial design, that more is coming in than we can take out. And I think how do we retain three years degrees and and manage that you America's four years degrees, Scotland, much of Europe. and I do think that yeah, I mean I think we might need more time to incorporate all the things we know they needed, plus the AI and things. I I think that could be a way of of bettering things.
Martin Bielicki: Let's hope the universities can can adopt in time. We definitely do believe we need new engineers and more engineers coming coming to the field. but yeah. Paul, it's
Paul Russell: Hmm. Yeah.
Martin Bielicki: great talking to you. Thank you very much for the conversation, your time.
Paul Russell: Thank you, thanks for your time
Paul Russell: Hi, yes. So I'm currently the programme leader for the masters in integrated industrial design at Loughborough. I teach heavily into the undergraduates as well product design and technology and industrial design.
Martin Bielicki: Awesome. Paul, let's jump right into it. So you've taught CAD for for years now and you've been in Loughborough where Chat GPT first landed and AI started really picking up. What have you seen change since then?
Paul Russell: well I would say even before even before, say chat GBT, I mean I was in the image space before the tech space. And I remember even like January twenty twenty one, you know, there's there was that demo from it would have been the first version of DALI, but even before that there was things like disco diffusion. There there were other things that just weren't doing anything, but but the second of that there was a demonstration where they were doing like, you know, chairs in the style of like an avocado. And that for me was pivotal in that, you know, you knew that was only early days and there were things there like the abstraction and the semiotics, the ability to embody it and to visually communicate it. And you think, well, okay, if you put that in front of a student or a designer, you know, they wouldn't bring that many concepts with that many much variety, as quickly. E even including the fact that, you know, things were touch and go, then y maybe one in five images was good. But the point being, you know, you could generate a hundred images and twenty of those would then be good, right? So that was really eye-opening and that moved fast. You know, I was in the betas for mid-journey in early 2021. And then things like stable diffusion dropped around the same time as Chat GPT. that was just really interesting. again, that kind of okay, maybe one in five images is good, but there's enough in there that it's valuable to kind of spend the time running this. and on the on the language side, you know, it was really interesting, even early days, GPT three, three point five, you know, ask asking it to kind of write, you know, five hundred words about some area of design practice. And there was enough in there, there was enough eloquence, there was enough understanding. And the and the the way that you looked at it, it's people people would look at outputs in the early days and compare it to their pit themselves and their peers. it's not that good. But you've got to understand there's eight billion people in this world, right? Probably seven billion of them are nothing about design. more than seven billion I'd say. it's it's that kind of narrowing of of of the kind of knowledge and understanding in that things are now kind of well, I don't want to get into semi s semantic debate about the word understanding, right? But the ability to communicate things and to perform kind of economically valuable work, it it is happening really, really fast. and we've got to keep on top of that.
Martin Bielicki: So you've seen the writing on the wall very early, right? Because I think
Paul Russell: Yeah.
Martin Bielicki: even now, I think engineers are we are very I guess skeptical species, you can say.
Paul Russell: Mm-hmm.
Martin Bielicki: And I think a lot of engineers, whether software or even more so hardware,
Paul Russell: Mm-hmm.
Martin Bielicki: still don't believe that AI can do their work. You know, whether it can do it fully, it's a question
Paul Russell: Mm-hmm. Absolutely. Mm-hmm.
Martin Bielicki: for another discussion. But why do you think you you've been so early to me. So why do you think you Like why did the skeptic in you not not kind of start kind of l sounding alarms and why did you believe so early that this will change fundamentally how engineers work?
Paul Russell: Yeah. I think I think in if you look at trajectories, you look for back from three ch chat GPT three back to two and you see that leap and you think, okay, even if the pace continued at this rate, right, and it wasn't exponential, it may maybe it is, I'm not that's not another another discussion.
Martin Bielicki: Mm-hmm.
Paul Russell: It it would only be ten, twenty years down the line that things would develop and and I mean I I I have been into kind of human intelligence since was very young. so I I and artificial intelligence as well. And and you can see If you can if you can codify things and see patterns, then you can kind of do it with a computer effectively. And a lot of what we do can be can be codified and you can you can see patterns in it, you can codify it, you can replicate it. and I think that's true of a lot of things that people don't want to accept are going to be
Martin Bielicki: Mm-hmm.
Paul Russell: true. And a lot of that is more human centered kind of things. But you know, talking about being an engineer, you know, I I straddle both sides. You know, I have an undergraduate in design, I have a masters from an engineering school. I teach kind of across I've worked across both, teach across both and You know, I I I I'm kind of foot in both fields and and I think you know, there's a lot of skepticism in design as well, but it's a different there are different kinds of skepticisms as as design engineers and engineers. and the reasons why, you know, people are adopting it and and resistant to it, kind of some are shared and some are different, but it's good to have that full spectrum and I think that that helps me as well to see like where it's gonna land and where it's gonna fit in because I I'm teaching across the board, you know, in terms of front end research. ideation, development, you know, the CAD rendering, technical drawings, you know, design for manufacture. And and I can see opportunities throughout the you know, across the spectrum for AI to be incorporated.
Martin Bielicki: Let's move to to the to actual engineering from from I guess from images, I guess. So how
Paul Russell: Mm-hmm.
Martin Bielicki: have you seen that shift play out of of AI starting to enter CAD and technical drawings, all the other parts of
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: the design process? And how have you seen it impact not only your students, but then the industry? Because I'm I'm sure you have ties to the industry at the university as
Paul Russell: Mm. Yeah. Yeah.
Martin Bielicki: well.
Paul Russell: It's very well, it's a very mixed bag at the moment, of course, because we have we have mesh models, like AI mesh models, and they are incredible and and I suppose that's slightly one of the criticisms I would have of of some design education is that if you can produce a fancy few renders at the end, right, and and a a G A, a basic G A, you know, you you you kind of jobs done, it doesn't go necessarily to the level of design for manufacture costing and you know, being able to do be produced at scale. So I think, you know, if you separate the two and then you look at you could you can you can do that today, right? You can make an image, make a mesh model, render it from a couple of angles, right? And you don't need that CAD model if you see design as being like, okay, this is like, you know, the the the the old notion of handoff to engineers, right? And in many ways that's not the case so much in the UK anymore. I think we are one of the most forefront kind of countries in terms of having that like link between design engineer engineering. But other countries, you know, there's still the concept of designer handing off to engineer and and then the engineer would have to model it properly, et cetera. but I think in terms of engineering, it's I think we're seeing early days of lots of ideas, lots of competition, lots of noise, lots of mess. And I think that when you look at what's out there, and that there are a couple of big companies, you know Bench is one, Leo is another. what's the other company called Backflip or something like that?
Martin Bielicki: Backflip, yeah. Mm-hmm.
Paul Russell: Backflip, right? Yeah. they're all looking at it from slightly different ways. I think Backflip's got an interesting approach. You know, they're looking at converting mesh models into parametric models. And that that's gonna have that will definitely have some benefit in that area. But you look at other companies that are trying to do CAD straight away. Some are doing it inside APIs, some are using existing kernels, some are some people are trying to write kernels. And there's loads of things happening. And yeah, it's difficult to say where things are going, but I do I am conscious that there are three big players, right? There's PTC, there's Autodesk, and there's Assault. And they particularly Audest, let's say, have have a a habit of buying things up when it kind of threatens them. And then they kind of you you arguably maybe they stunt innovation sometimes, or they integrate it fairly well, like they did f with T spines, for example.
Martin Bielicki: Mm-hmm.
Paul Russell: showcase and fusion and things. But other things they kind of buy it and and let it go. and DeSalt have, you know, they are not I don't know they have a hesitance to buy things because I don't know they feel like they can just do it better. And then PTC, of course, have just bought OnShape, but they're also par in parallel at the moment, trying to write like AI for OnShape and AI for CREO. They're completely different kernels, you know, they're like doubling up their own work. And I think there's a lot of mess. And I think that what what you see you're gonna find there are limitations not only in terms of the level of access that companies will give to the API, but I think just in terms of even kernels, you know. I mean if you think about CAD, like I I've been thinking about this recently. You think about like SolidWorks, right? So it uses Parasolid, but so does Shaper 3D, so does Plasticity, so does et cetera. And and if you think about like the kernel, the kernel is like a model, right? And then on top of that, you have the API. And the API is like a harness, right? And when you when you think about things that way, you can see a lot of the problems aren't necessarily with the kernel, they are with the harness, right? And you could but you're gonna have limitations where the haun the ha the model and the harness aren't being made by the same company. That level of integration will be limited. and I think the real potential then is for players to come in. And build something that is a modeler from scratch. You know, it can model, it can do assemblies, it can do simulation, PLM, but the whole kind of works, right?
Martin Bielicki: I think that's that's the golden really outcome. I think the difficulty
Paul Russell: Yeah. Mm-hmm.
Martin Bielicki: is the first thing that corporates tell us currently when we speak to them is I don't wanna switch away from CATIA I don't wanna switch away from Siemens NX It's the first sentence we hear. So that's
Paul Russell: Yeah, yeah. Mm-hmm. Mm-hmm.
Martin Bielicki: the difficulty, right? They're obviously imperfections and these
Paul Russell: Yes.
Martin Bielicki: softwares are obviously not optimized for agents to use. They're not AI native software
Paul Russell: No, no. No, no.
Martin Bielicki: in kind of whatever definition we're gonna use.
Paul Russell: Mm-hmm.
Martin Bielicki: But also we have to meet the industry where where they are. And kind of
Paul Russell: Yes.
Martin Bielicki: the the cost of retraining a whole workforce of BMW to, you know,
Paul Russell: Mm-hmm.
Martin Bielicki: a novel AI CAD, as much
Paul Russell: Mm-hmm.
Martin Bielicki: as it would be probably greater at some point. I do believe it happens. When does it happen? I think it's
Paul Russell: Mm-hmm. Right.
Martin Bielicki: not for another couple of years at least. And I think that benefit would have to be so clear for them to to actually undertake
Paul Russell: Yeah. Yes, yeah, yeah.
Martin Bielicki: that that motion.
Paul Russell: Yeah. And then on the on the other side you've got I mean, I have seen more new CAD programs and attempts to build kernels in the last two months than I've seen in the last ten years. That is happening really quick. There's a lot of companies out there making parametric modellers, basic ones, you know. But so some are using O C C T but some are building like kernels. some are making kind of brand new surface modellers from scratch, you know, they replicate things like Rhinos or alias even. and I think you'll see a lot in that area as well. And that's another interesting part because you know we talk about mesh, we talk about parametric. But there is also the kind of okay, are we talking about you know simple you know feature tree parametrically modeling parts, or we or we're talking about surfacing because there is a difference there as well. And think of course I think we're we're we're probably yeah we're definitely less than a year away from you know the standard kind of e exercises you get in university like here's this fully dimensioned three view drawing model it in CAD, right? One click to get that in a in a step file. That's well far less than a year away. And when we
Martin Bielicki: Yeah.
Paul Russell: can do that, you know, how does that impact how we teach CAD? If a student knows I can just one click it, well where's the motivation there to kind of learn it? But they need to learn it to have the basis to do more advanced things. And that's going to be an issue
Martin Bielicki: Because
Paul Russell: we face.
Martin Bielicki: yeah, I guess that's my next question. How I think that's a broader question about academics. Like, what do we teach now that AI is here? But of course we we
Paul Russell: Yes. Yeah.
Martin Bielicki: should focus on on design engineers and and cat in general. How do you approach this? So how do you approach teaching your students to enter this new paradigm well
Paul Russell: Mm-hmm.
Martin Bielicki: with the necessary skill set of the kind of bedrock
Paul Russell: Yeah.
Martin Bielicki: skills you need to be an engineer? While also not them them not feeling like they're they're doing manual work and they're kind of twenty years behind what's what where the industry is.
Paul Russell: it's not threatening CAD just yet, but I think I was I think there's still possibility by October, for example, that the one that they have inside Fusion is up, running and pretty functional in terms of you can use natural natural language, you know, build this block, fill it these edges, and it will understand, you know, if I have like a block and I say fill it the vertical edges, it will have enough semantic understanding to pick up what what that means. And I I think I would say that probably by October next year we'll have a lot of students coming in with fusion experience that have experience with the AI as well. And they probably will see quite mm that there there definitely will be limitations in in how complex it can be. But yeah, I mean the the the difficulty there is there's another factor in education in that there's quite a lot of students. I mean we have you know, we have say you've got five hundred students, you know, you'll have hundreds that are fantastic, great, brilliant. Th there are students that just wanna do a degree, right? They just wanna get through the degree. And I think those students will be able to do that with AI, even with CAD by next year, next October intake. And and how do we manage that? Because then what are we you know, are we delivering an education or are we allowing them to kind of jump through the hoops to kind of do enough work to kind of get a degree? And we nobody wants that. You know, it's not good for the students, not good for academia. It's not good for, you know, people that that, you know, are arguing that there's too many graduates already anyway. And you know, there's there's a lot of factors involved there. but how how do we then Yeah, I mean I I think it the same with a lot of other stuff that I'm doing that, you I do a lot of I mean, I've taught for years key shot and blender rendering, right? Or really high end stuff in Blender. and we s I you know, Gemini Nana Banana came out and then Veo and you're like, my god, like this is insane. The the the speed and the accuracy and the context and the realism, even if you hash it at five, ten, fifteen times, right? It takes dozens of hours to make work in Blender that that's not even of that level of quality. And that's something I've had to face already, right? So as an example. and I I had to kind of bite the bullet, you know, I took years to learn Blender. I enjoy Blender. Actually, and that's another subject I'm sure we'll come on to, you know. What what about the enjoyment, right? Like I enjoy using CAD. but I ha I I'm not gonna hold back and say we must use KeyShot to make subpar renders because I I don't know, it just doesn't make sense. You look at the history of of what we do, you look think about drafting, right? I I my first design engineering job. There was a draw full, like a planar chest of like A1, like hand drawn, hand lettered written out drawings. Nobody does that anymore. We can't nobody can really do that anymore. and we don't consider it time effective, effective use of time. It's not cost effective. It's not, you know, it's not robust enough in that, you someone spills coffee on it, you're in trouble. so things things do change and and we have to evolve and adapt and and I'm trying to kind of keep on top of that. And actually not only that, but be slightly ahead of the curve because You have to understand as well that most of our students do a placement. So a student entering this year will leave into the workplace in 2030 in four years' time. I am not training a student for work in twenty twenty six. I am training a student for work in twenty thirty. And that is a really critical thing that that most educators haven't quite gotten to yet.
Martin Bielicki: And that is also so hard that you to try to predict that future, right?
Paul Russell: Absolutely. You yeah, and exactly and so and sometimes it is i it's a bit of a gamble, some of it will be right, some of it might not be exactly right.
Martin Bielicki: Yeah.
Paul Russell: but you have to do that. You can't put people out into industry that, you know, are are designed you know, prepared for industry as it was when they entered.
Martin Bielicki: On that topic, you mentioned that AI can make good designers better, but it it won't do much for mediocre or bad designers.
Paul Russell: Yeah.
Martin Bielicki: Why do you think that and how do you think that that plays out over the next couple of years?
Paul Russell: Yeah, I mean I I wouldn't say only AI. I mean perhaps it's like technology like like having access to like renderers, like being good at rendering, is better than for example, you know, old days of Photoshop rendering and before that marker rendering, before that chalk and pastels, right? So technology helps people that that kind of master them to get even better. And and the same with CAD, right? What's the alternative to CAD? You know, your hand drawing, technical drawings, right? you know, the speed, the efficiency, the kind of ability to I mean s some things, you know, that that are very difficult to draw, can easier to CAD model sweeps, lofts and things, right? so th not just AI, but CAD helps getting good at CAD helps good designers get even better, right? And if you just kind of learn the basics and only pick up the basics, you know, we w you see students that graduate only being able to do extrudes and fill it, right? That that happens, right? And some of them, to be fair, get to relatively complex forms. Granted they're only using like extrude boss, extrude cut and fill it. But they get through it, right? They get jobs done. But being able to learn, for example, you know, more advanced tools, both in solid and surfacing, does enhance their work. Same with AI, right? But if you engage with it, I mean I've been teaching AI for many years now since 2022. And I suppose it it's no different than anything else, but but you you can learn it passively at a surface level. You can kind of tick the boxes or you can really engage with it. You know, a student could and an interesting one, right? So it ca it takes for example, some of the modules this year, right, using flow, it's three hundred percent free, so there's no cost to it. You open a web page, you put a text prompt in, you get four images back, right? It takes twenty seconds. You could do that, you could spend five minutes, right? And you could have hundreds of images. And even better than that, you know, like I was saying, four years ago, one out of five images would be roughly usable, right? Today, you can make one image with five design concepts on it. Of which a handful are usable. And you're making four images at a time, you look at the scale of that, right? And that'll only increase the the capacity to do things like that, to get ideation, inspiration, right? but you know, one student might do a hundred images, one student might do f five and feel like they've ticked the box, right? And then it's that level of engagement. But it the more that you use it, the more that you learn from it, the more that you learn to work with it, the you the more that you learn how to use it, how to engage with it, how to share agency, how to co-design effectively and and that's gonna happen a lot in CAD too, right? So if you think about CAD, like before it existed, a lot of the kind of idea was that it was computer aided design, right? The computer would aid you to design. But we've used it for, you know, four years as a drafting tool, right? It's a digital drafting tool. It's not it doesn't aid you in any any way. There's a plug in for SOLIDWORKS like DFM Advisor and and it's it's very primitive compared to what we could already do, right? And Autodesk Fusion has one and they're working quite heavily on it. But I think that is where the future goes as well, right? You might model a part and the the AI might be constantly monitoring that part and saying, No, actually that wall thickness needs to be thicker, thinner, you know, the ribs need to be thicker, thinner, you need a boss here, right? And people will have to learn to share agency with, you know, the computer or, you know, the the the AI inside the CAD, just like they will, you know, things like research, you know, front end research, secondary issues that is you know ideation inspir you know getting inspiration you know all kinds of things but it will happen in CAD too and tech like technical
Martin Bielicki: I think
Paul Russell: drawings as well right I mean that's going to be an even easier one right you've got drawing and it's gonna say well you're missing the dimension here or that doesn't have the right tolerance or you know th these things will happen and people will have to get comfortable with that.
Martin Bielicki: What do we generally believe is that, kind of on the topic of enjoyment as well, but kind of I think even more so impact, is that broadly engineering is going where software went. Where essentially now engineers and software
Paul Russell: Yeah, yeah.
Martin Bielicki: engineers essentially manage swarms of agents and essentially ride the system that directs the agents.
Paul Russell: Mm-hmm. Yes, yeah. Mm-hmm. Mm-hmm.
Martin Bielicki: And that essentially what we've seen kind of and kind of how our our software engineers
Paul Russell: Yeah, yeah.
Martin Bielicki: code, the enjoyment is very much there. It's just the pace.
Paul Russell: Mm-hmm.
Martin Bielicki: has picked up so much.
Paul Russell: Yeah, yeah.
Martin Bielicki: And it's it's basically the limitation seems to be the the cognitive abilities of the humans. Essentially how many
Paul Russell: Mm-hmm.
Martin Bielicki: agents can you manage at the same time? Because technically you do infinity. But then your your context
Paul Russell: Yeah. I think it's gonna be a bit slower yeah, I think the pace will be a bit slower for engineering. But I I do I have followed, you know, heavily like coding and programming, you know, how how people are accepting it, adopting it, the resistance and you know, there's resistance, resistance, resistance until the the the system's got to such a capable level and lots of people now like, Okay, it's changed, right? There's nothing I can do about it, right? It's yeah, it is what it is.
Martin Bielicki: There's no question now. There's no question. Like two years ago we had we had skeptics and now I I've I very rarely get any s skeptics on on AI and software.
Paul Russell: I honestly I I've been very skeptical of AI CAD myself, but I've seen enough now to know to I can see where things are going. I can see the patterns, I can see the trajectory, I can see, you know, how you can emulate lots of things. So it will it will happen. but Catia
Martin Bielicki: Wa
Paul Russell: Catia and Siemens, sorry, are are already quite they're a bit down that line of agents, right? They already have abilities to kind of request and and constrain things and get results in CAD, right?
Martin Bielicki: I think they do. Yeah. Catia's Catia, for example, Dassault are working on on their Leo agent, for example. They've got a couple, I think three agents. I forgot the other two's names, but but Leo is the one doing doing CAD.
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: for sure it's getting better. I don't know the exact capabilities, but yeah, it's getting better and better. why do you think the progress is gonna be slower in in engineering? Do you think it's more so a technological block or a cultural block?
Paul Russell: Yeah. I mean it's it's another dimension, isn't it? I mean I suppose, you know, even with programming to an extent you can work in in three dimensions. You could you can design games, right? That that are in three dimensions, applications that display graphics in three dimensions. I don't know, I think that there is more to it in terms of materiality and I think, you know, simulating a piece of software is relatively simple and I think I mean it's it's more data, I suppose, in physical side, you know, everything must have typically kind of some level of i it required physical strength, you know, weight limitations, heat transfer, you know, there's gonna be a lot and and costing as well, right? So I mean a lot of, you know, what I teach when I teach design for manufacturers, okay, you've got like you've got like cost, you've got strength, you've got aesthetics, you've got like other human factors, like is it cold to the touch, is it comfortable to hold, you know. And the like the intertwining web between all of these things is so complex. I do think it's a level of complexity greater than than programming. And I think what you'll see, I mean I think you'll see things like if you look in the large language model space, you've had people that trained initially like coding models, right? It did coding, you did like ones that did like natural language literature poems, whatever. You had ones of people that would do it for certain different things. I'm not gonna use the term of general intelligence here, but you know, you see things over time like come together. I think you will see things in isolation in engineering physical design. And I think it'll be a little bit of time until they start pulling together to create kind of models that are generally capable in in engineering.
Martin Bielicki: Mm-hmm. Let's let's pull that string a bit. you've got a a course about the ethics, the ethics and aesthetics of generative AI in design practice. Could you tell us more about that? Kind of how how ethics and aesthetics are intertwined intertwined with generative
Paul Russell: Yeah.
Martin Bielicki: AI.
Paul Russell: Yeah, I mean so I think I think what a lot of it was is you have a lot of people that say, these these images are great. you know, the argument is are they good and are they not good? And then you have like people that are arguing, you know, are are they is it ethical or responsible to use AI, training data, etcetera? is it unethical? And I think those conversations we've maybe I had in kind of isolation, you know, they're really you know, to get everyone to kind of some level of like understanding of everyone's position. You know, I I'm in the module I'm happy for people to say and and and more so in graphic design, because a lot of students take it from graphic design as well. You know, they say, I I don't think the results are that good. I th I think I'm better I would generally disagree, but that's a different subject, right? But they're allowed to have that opinion, right? And and that's fine because the point is that they leave the module with an informed opinion, not that they leave in harmony, right? Agreeing with one another or with me. so I think, you know, you look you know, some people can look at the images and they could say, Well, they are beautiful, but actually is it you know, ethical and they could say, Well, you know, it's unethical. Sorry, sorry, the images are terrible and and but it, you know, it is ethical and and I think there's a kind of crossover there where everyone's having kind of two arguments across one another and I think they are more linked than you realise. I mean ethics is sorry, aesthetics is a form of ethics and ethics is a form of aesthetics. They're not two different things, right? So my my real plan with that was to kind of get people generating content, evaluating you know only not only the quality of the content, but using a variety of models, comparing the models to one another. and then also I teach environmental kind of impacts of AI, both on the hardware side, you know, that we need more graphics cards. We have to force people to mine the Congo and and other countries, and that's not great. you look at the socioeconomic side, okay, so okay, it can emulate work and maybe replicate and and kind of replace work eventually. Where does that leave us? What do we do? Do we reskill? And and also, you know, does it reshift the location where people work, right? So maybe there's lots of people working in the UK. You look at say data centers, they were outsourced, you know, Malaysia, India, et cetera, et cetera. You know, if someone if their only duty is to watch agents and kind of manage agents, do we need to be paying them, you know, high salaries? you know, all all these kind of questions that that that need to be had. and also like I it's an opportunity as well to bring in intellectual property, copyright, patents, design rights and things, 'cause that's not covered too well generally in education.
Martin Bielicki: What we're seeing in many areas of of work done that is a huge bifurcation of the top ten percent getting paid more than ever and are more valuable than ever, and
Paul Russell: Mm, mm, mm, mm.
Martin Bielicki: the bottom ninety are trending downwards.
Paul Russell: Yeah, yeah.
Martin Bielicki: my guess would be the same happens in design and engineering, as
Paul Russell: Potentially.
Martin Bielicki: as you can scale up the impact of any and like the best individuals somewhat in infinitely.
Paul Russell: Yeah, yeah. Potentially yeah, I th I think it's two conversations there. Th the one the one thing that you that people really need to be aware of is a lot of the people that are taking a hundred million pound a year jobs to train AI models are doing so and the salaries are so high because they know there won't be any jobs in programming in a few years, right? As soon as we have self kind of training models, they know there won't be jobs, so they have to take those salaries now to kind of be future proof to to to have money to kind of get through, right? So we we know that's happening and and so it might it might look like okay, some people are getting paid loads of money, but I think they know as better than anybody that the money is high because they are training the models to the highest levels to replicate the work that even they do. The flip side is you know, a scenario where we have global power shortages, GPU shortages, they'll put another world war which isn't off the table at at all. And you know, we lose these systems or we lose the the access or all kinds of things or or or politics, geopolitics that prevents kind of access to models. you know, I think if we do get to a situation where models can do everything and nobody thinks that anything's worth learning and and training on, yeah, we are kind of in trouble when if something happened that prevented us from kind of having access or using models.
Martin Bielicki: Yeah, and I I think that's where the the whole topic of sovereign AI owned by continents or nations or groups like the European Union, that's where this becomes becomes interesting.
Paul Russell: Mm-hmm. Yeah, yeah.
Martin Bielicki: Let's let's put our creative hats on for a second and let's try to imagine what does the job of a design engineer look like in five years? Well because obviously
Paul Russell: Five years.
Martin Bielicki: as you as you said, you you're just training your your students, Carney, for a century twenty thirty.
Paul Russell: Mm-hmm, mm-hmm. Mm-hmm.
Martin Bielicki: what do you think that looks like? I know you have a an interesting post on LinkedIn about how a design agency of the future can co could ta theoretically alterate more and more of its work, trending to like nearly a hundred percent in the coming years. How do you think that that plays out? Kind of what what are you telling your students and how are you advising actual design engineers that are currently in the market to to prepare?
Paul Russell: Yeah, I mean one one of the things I did, you know, a few years ago is I bought Blender and Rhino into the courses because I knew from the get go that e everything that was gonna happen with AI was gonna be in those programs first for a couple of reasons, right? Blender's hundred percent open source, open kernel. but technically Rhino is an open kernel as well. So Rhino the Rhino kernel is open source, so a lot of people write for that and and tinker with it and play with it. It's affordable. So a lot of you know the stuff I've done in recent years, you chat GPD, write some code, write some grasshopper dis definitions, do all kinds of things in Blender, right? And that's I'm not I'm not even touching on MCP or any of that kind of level of of automation yet. so one of the things I did was to say, look, you know, you don't don't understand CAD as parametric modeling, don't understand CAD as SOLIDWORKS, you know, taking things away from rote learning or skill learning or tool learning to higher level understanding of geometry, of computation, you know, things that you know that if you're gonna be mod watching kind of agents and things that you're gonna learn need different kinds of knowledge and understanding. so I think that's part of what I d I've done there. Now design engineering, I mean the things like technical drawings, right? I mean I as probably with all people, you graduate from product design, so my background is like product design, and you get sold this kind of dreams, you'd study and like you get like freedom, you get crazy projects, you get like design freedom. But the real world's not like that. and and you go into industry and you're like you expect to be sitting down all day designing CAD modeling and all these kind of things. And you f you realise, okay, I'm spending as much time an Outlook as I am in SolidWorks, right? But then there's a big chunk of time where you like engineering drawings and and even like, you know, CAM programming, CNC programming. Like I used to do a lot of laser and punch and things. And it's like they the the job's not that sexy, right? Some of it a lot of it is just not. But I think you will see yeah, I think this is the kind of worrying part, is you'll see a lot of that drudge work be automated and the drudge work is more easy to automate. And I think what you'll then see I mean there's two ways you can see it, right? Everyone just reskills and gains these new skills and suddenly the like the world needs more stuff. And I don't think that is going to be the case. and the flip side of course is that, you know, we need less people. You have five engineers in an office, right? You take The 30% of their job away at the storage work, right? You 30% of rest people. You drop down to three people, right? For example, I'm not saying this is going happen that the values are just like guesstimates, right? But that that's very likely to happen in many fields. So the purpose of what I'm doing as well is to say, look, you guys are getting those skills more so than any other degree I know. You are more prepared, more kind of informed. You have not only like skills and knowledge to like use tools and and do things but also, you know, I'm I'm doing that ethics and aesthetics. We're looking at philosophy, responsibility, you know, maintaining kind of design rights, human kind of touch, the whole kind of works really. but if you put me on the spot and said what does the future look like, I would say we are gonna have less jobs won't disappear, but if you take s a certain portion of people's job away and you take that away from everyone, you need less people to do jobs and I think that does seem to be where things will go.
Martin Bielicki: Yeah, I'm I'm curious how how it's gonna play out in terms of the market expanding. 'Cause just like I I kind of the the software market I make a lot of software parallels here, but the software market has expanded a lot after the I just because we now it's
Paul Russell: Mm, mm.
Martin Bielicki: now feasible to build so much software and people
Paul Russell: Yeah.
Martin Bielicki: can can now build one off apps here and there on their own because they can now you know vibe code, if you will, their own apps. I'm curious if if that that same will happen in engineering, where essentially if making stuff is faster and easier,
Paul Russell: yeah. Hmm.
Martin Bielicki: maybe it's actually now we produce more. it's of course a bit different. There's a limitation on manufacturing, producing. That's harder to access for a normal person. But yeah, that's an angle we also are looking at.
Paul Russell: I I would agree that there's gonna be definitely a democratisation of people that can go from an idea to a physical product. I don't think the world needs more stuff. I think there will definitely be a period where more stuff happens. And I think you know, you look at things like Terrible Places, Alibaba, Timu, that kind of route of kind of getting stuff out there. I think a lot of that'll be AI generated in in time 'cause a lot of it's kind of cutesy stuff, it's kind of, you know, bit of graphics on the world you know. Stuff like that, teddies and And all kinds of things, lots of that work will go to AI. on the high end engineering stuff, I mean I mean it's a sad reality that we live in, but right now we need more military equipment. Europe needs more, America needs more. you know, we we need that and and I think that there could be a lot more work in those areas in the in the next few years. And I think that's the kind of terrible human nature that that will be sustainable kind of careers to be in. But do we need More I mean if there were more planes out there would flying to America be cheaper? I don't know, I don't we need do we need more planes to we definitely need better planes, but yeah. Stuff like that.
Martin Bielicki: It's it's a tough question, right? It's a tough question. I think we we spoke to a lot of automotive companies recently and they said, for example, that the pressure they're getting is that China's producing a new car every year, like as if it's a new phone. And then I have
Paul Russell: Mm hmm. Mm-hmm. Mm-hmm. Mm-hmm. Absolutely, yeah.
Martin Bielicki: to pick up their design iteration, their time to market, just to really service that demand and be able to sell those OEMs without have to deliver a car every year.
Paul Russell: Yes. Yep. Yep.
Martin Bielicki: That's an example where I think we'll get more stuff. Whether we need it, probably a different topic.
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: but yeah, but even beyond military, we definitely have heard from across the industry that broadly in engineering there's more so a shortage than an oversupply of engineers. And and now especially with the cohort of
Paul Russell: Yeah, I mean that's that's a really interesting one. I mean for years in the UK you see it all the time not not enough people to fill engineering roles and you see things like you know engineer applied to loads of jobs, never got a kind of job in that. I don't know, it's a weird one. I mean so many of the jobs out there, like you go on like indeed or something, right? Most of the jobs are fake because they're recruiters and they're kind of fake jobs to get people's details on a system to spam them with jobs if they do come up. so is it really hard to determine real things like that. But the the reality is that you look at, for example, fantastic great universities like Loughborough, right? And say, you know, people who actually go into design engineering jobs directly, not like loosely related. It's probably about thirty percent, you know, you know the thirty percent that they go into, you know, loosely related roles might just be a CAD rendering or whatnot, right? I mean even like the engineering school f I mean you and I mean the statistics aren't really great, but the best that we have, you know, it might be fifty percent of say the engineers that get a job, right? But these are top top class kind of graduates from a great course, great university. So I don't know the notion that there aren't the jobs out there, I don't I don't sorry, there aren't enough engineers out there to fill the roles. I don't really agree with I know that that that the news says that and the statistics say that in the media and whatever, but I don't know if it's true. I don't know if it's true. Yeah, but but the the weird thing
Martin Bielicki: We hear it from from the industry though. I mm maybe there's something around from
Paul Russell: is, yeah, that they say that, but then they they kind of a lot of them have they refuse to deploy certain people and like For example, like an injection molding job right club and they like, I I'm only employing people who have got injection molding experience and that kind of mentality and then they don't recruit anyone,
Martin Bielicki: Mm-hmm.
Paul Russell: they're like, there's nobody out there. I mean that that is just for me is like absurd, right? same with machinists, right? Yeah.
Martin Bielicki: Maybe yeah, maybe there's a mismatch. Yeah, maybe there's a mismatch between what they're looking for, what they think they're looking for, and what's out there in market.
Paul Russell: Yeah, I mean no serious sometimes it's about it's like they don't want to take any responsibility or onus on for like training or development. They want people to come in the door, be able to able to do the job and I think that's an issue.
Martin Bielicki: Ready, yeah. Yeah, the I I can definitely see that. I think that there's probably something something something around there. Awesome. Paul, I've got one last question to close this off. if a seventeen year old came to you now and said they wanted to be a designer,
Paul Russell: Mm-hmm. Mm-hmm.
Martin Bielicki: knowing all that you know, what would you advise them to do? Do you think they still need university? Do you think they can self teach things? Do you think they need to learn cat bottling from scratch on everything? it's a very broad question, but I'm curious to hear your take.
Paul Russell: Yeah. I suppose like my answer to that question would also depend on like how fast can universities evolve and adapt, right? I if the universities could adapt and evolve quick enough, definitely like university brings way more than the skills, right? And the knowledge, it brings the experience of like working in groups before you have to deal with like you know, you have experience working with people in groups and one might be not pulling their way or annoying and that's like real work, right? That's like that's like the real world and it prepares you for that. And there's so much more, of course, you know, like a degree. has a like an academic element to it, you know, you learn theory and philosophy and things like you learn to like a higher level than than you would just gain like the skills and knowledge. But I mean if I was seventeen today, w what would I do? I would like I mean I think it's I suppose the answer's true regardless, but learn materiality, just look around you, plu like use materials like wood, metal, you know, polymers, like look at what things are made of. Look, you know, gain Like because they're kind of there, yeah, I think in some ways like so my research is mostly on design evaluation now using AI to like enhance and augment like concept selection evaluation. And I think that's the kind of bottleneck right now, but I think that bottleneck will go. I think the bottleneck will be like can I get it made and manufactured? And I think the the design for manufacture knowledge and the material selection and and those and the costing kind of end of things are gonna be like the last bastions, you know what I mean? Like those last kind of real valuable skills. I mean AI will definitely get there eventually, but I think I think yeah, if you look at what's gonna be valuable, economically valuable for employers, the longest and the hardest for AI to kind of overcome, it is like the f feel, the physicality of the world. and I think that's what designers really need to be on top of now.
Martin Bielicki: Quick bonus question here. Do you think universities will be able to adopt in time? Because from my understanding, universities take three years to adopt a new new course or a new curriculum. And with the pace of AI change, three years is a whole era, right? It's a completely different paradigm.
Paul Russell: Yeah. But yeah, I don't know. I think if you look at higher education, it's like a slow moving kind of system and bureaucracy and things, you know, that's just d nature of it. And and it's the same thing like big automotive companies, right? Now I've I've heard a lot about, you know, companies that want to adopt Blender for modeling and rendering, right? And the the amount of kind of hubes they have to jump through for for what seems like a basic thing, it's free and open source, right? What's the problem, right? There's there's no cost involved, we don't have to cost it and all these kind of things, but they still have massive like systems and layers and and years of kind of getting things in. I think AI will be the same for a lot of companies, right? The b the big automotive companies. They'll they'll be very resistant. They'll have to do, you know, internal validation of security and all things. Can I can universities move quick enough? I don't know. I kinda I can say I don't know. I wouldn't say definitely I wouldn't say definitely not. But I think and I hope that AI does you know like the the kind of course kind of reevaluation, redesign, right? That that is definitely, yeah, you're right, like three, four years, years ago, I think that has changed. I think that things are more I think things are monitored more regularly now and and kind of redesigned, redeveloped, you know, pushed and pulled. I think if you look at the history of design education, things have always been that way. More things come in, some things kind of have to go out. But I think we have a problem in higher education now, in design education, industrial design, that more is coming in than we can take out. And I think how do we retain three years degrees and and manage that you America's four years degrees, Scotland, much of Europe. and I do think that yeah, I mean I think we might need more time to incorporate all the things we know they needed, plus the AI and things. I I think that could be a way of of bettering things.
Martin Bielicki: Let's hope the universities can can adopt in time. We definitely do believe we need new engineers and more engineers coming coming to the field. but yeah. Paul, it's
Paul Russell: Hmm. Yeah.
Martin Bielicki: great talking to you. Thank you very much for the conversation, your time.
Paul Russell: Thank you, thanks for your time
