Name: Brainstorm Design 2023: How Ai Is Transforming Architecture
Uploaded: 2023-12-07
Duration: 23 min 34 s
Description: Tim Fu, Designer, Studio Tim Fu He Wanyu, Founder, CEO, XKool Tech Moderated by: Wang Fang, FORTUNE CHINA

Tim Fu, Designer, Studio Tim Fu He Wanyu, Founder, CEO, XKool Tech   Moderated by: Wang Fang, FORTUNE CHINA

Brainstorm Design 2023: How Ai Is Transforming Architecture

Conferences

So Tim, we have heard a lot about your great work. Let us start, let you, I mean, let me start by asking you to give us a really quick run through some of the projects that you've been working on. Mm Right. Just to change things up, give a little five minute uh demo just to show through the some of the stuff that we do with artificial intelligence. Um My story started when I was still working as Al Hadid architects. Um It is here, I learned how to take complex geometries, how to work with them and eventually turn them into build objects and architecture. So with that background, um I started looking into the first program uh mid journey and the early days I explored a lot how to um figure out how to control the A I and design and potentially create things that are interesting and architecturally significant uh in terms of uh its pedagogical values. And uh I started sharing a lot of my work on social media and I think things kind of took off on its own. And I delve really deep into the exploration of A I and also using social media to share my contents and uh showcase its power and its prowess is a technology today. A I can do a lot. This has been one year of progress since I started. And with a lot of my work, I like to incorporate my infatuations with organic geometry and a lot of futuristic I would say and contemporary styles and finding ways to merge them where you can use lines and curves manipulated by the designer using the algorithm in order to get what you want to distill an information and idea that you had in your mind into something that looks potentially feasibly construct. With A I. We can explore the merits between tradition and modernity. I like to explore traditions ranging from the West, Middle Eastern as well as far East. So that's mostly through my journey. And then I discovered look X slash uh X cool in this domestic markets and we have our, the CEO on you here as well. So um interestingly what Keith showed earlier was actually a demo from uh her platform as well. So I think it's very interesting. We live in a time where um a lot of the technology converge and all of us architects find the same tools when we find it useful with. Look X. Though beyond just sketching, I like to investigates other methods of uh using A I for the design process. You can crumble a piece of paper, ask for it to prompt a Frank Gehry architecture and then do the same with potentially Zaha Hadid or in the style of Daniel Lek. I also try to find ways to incorporate sketch as well on my side because it's a little more human touch and I try to pump it into the architecture styles that I look forward to. When I'm designing through the traditional design style. Sometimes you could get inspirations from nature, organic forms and geometries and turn them into ideas of how we can potentially build the future of our cities. So today, I have left Al Hadid in order to pursue my own career in my own office because I took a big leap when I found out about A I and I believe we have to have a very radical way to recreate a new process in the architectural design practice to really heavily investigate the potential use of A I in all domains. This goes beyond just making pictures, how we connect algorithms designs with the power of A I to create rapid ideation, how we use A I to generate program distribution and using the visual A is in order to produce a qualitative results and feedback. It's about the immediacy of doing these things, manipulating design decisions and allowing the final results to take form in front of our eyes so that we can make these informed decisions more accurately. And I have various projects around the world that I'm in the process of engaging in. Although marked with NDA, I can showcase some of the concept designs that led on to further developments and this range from urban scale architecture to uh more potentially public cities and plazas. So to conclude, I would like to say that the future is uh really bright with A I. But really, we're just at the very start of it at the Nascent. So let's try to build a future together and see what A I and human can do together as co creators of our future. Thank you. Thank you, Tim. As Tim said, um the wonderful app that has enabled some of the great work that Tim showed us is called Look X. And it's exactly the same app that Keith used in his earlier session that turned his sketches into very vivid images of buildings this morning. So we have the creator of this app. M He, you right here with us. So when we have seen how sketches and image prompts are fed into the app, we've also seen the, the ending results out of the app. So can you show us what's happening in between, what are the thought process so to speak inside the app? Yeah. Um I also gonna show a few slides to um basically uh like um sharing my ideas. Firstly, our company called the X Cool. We have two kinds of products. One is called the LX, which is generating ideas, concepts, virtual ideas. That is for creativity and another part is for more engineering. So we can do the uh for instance, the compliance, the calculation, the every details for the design. So this is the basic two parts and the look a here can do three things, generation training. So meaning you can train your own models and do cos sharing so everybody can learn from each other. So talk about the base is the LX is based on the diffusion model. But with our own data and trained with uh computing power for quite a long while. And also like updating the algorithm. And then we have this, that's the result. Um And this is the base I can share how we do it. There's a three ways to do it. Firstly, we can do text to image, it's a similar to the journey, maybe everyone in our have tried. But here we can do some very simple labels. Like here you can select your labels more for the architectural design. Like the terminologist will usually use and the A I will elaborate the prompt for you. So you don't have to write the whole sentence and the A I can generate ideas for you. But those ideas are more like um searching for every possibility but then be more precise as architects to control the idea, we will have other um like ideas to do it. But the, the base is we have this algorithm model as the base people can control the files So we have a small like a low Laura model, low ranking adoption models. So people can select and merge their files. For instance, here is the hands skit model and the section model. And this algorithm models, you can combine them together and to create your own file out from it. So all those are based on the text to image. And the second part is we can do image to image and this image is more like referring to what we want to precisely control. For instance, here you can see uh we can upload our images like the hand sketch, a team show or the crumbling paper or from a picture from the nature or from the massing model whatsoever. And with those images, you just upload it and they give a reference image, this is like a one shot learning and then we got this result just in Yeah. Second, you can get it immediately and you can slide like in between and see is this precisely um like to your idea or not? Of course, you can also do other sketches or different ideas to try out everything. So uh our platform is enlarging uh like everyone can like uh show the idea even they can just uh scratching something. Um Yeah, from the the real world. And the last yeah thing is that we can do editing. So based on image, if you want to add some part for instance of this here, you want a part of the facade. You just um yeah, brushing out the area and ask the machine to regenerate this specific area. Then you can get a new sub out front for here. You can get a wooden facade instead of the previous glass facade or you can also do the massing to change it. So these are three ways to elaborate the generation. Thank you so much. It's wonderful. So Tim next question goes to you. So in your slides, we show you showed how you use crumpled paper as a picture prompt. I mean to feed into the app so that it creates images just like the crumpled paper. So and I think in a, in a recent article, you said something like um when you design with A I whimsicality and intention can both play a role. So can you use that crumpled paper as an example and tell us ela elaborate a little bit more on, I mean, the two elements working together? Yes, I think now in the design process with A I, um there's this idea of a dichotomy between what you can control and what you allow the A I to provide for you. So I think as a designer and an architect, you want to have a differently, a degree of control from your own design. S os in your perspective. But then uh where A I can come in is also and provide alternative options, provide uh how they, how they hallucinate the details in ways that you would not have preconceived. Um So I would say that the economy is about actually allowing a degree of that wins and Cali to take over. So that as a designer, although you are sure of your overall vision, you want to give A I an opportunity to express somewhat its own degree of creativity so that uh you become the overall curator of that process, but you almost allow A I to become like a co creator. Uh Another designer that's in your team to work together. Are you often surprised by what A I gives you? Yes. Yes. Uh Sometimes in a very bad way as well, it all depends, it's like rolling a dice. And so there's a saying that we should not rule out hallucination completely because sometimes great ideas come from hallucination. So what we've all seen what A I can do so tell me what A I cannot do at this stage for an architect. Yeah, I think the limitations of A I is uh trying to engage in every process going from design to construction, right? Because as many have already clearly stated that uh the design process is goes from the early concepts all the way to the development and the production and construction. And I think it's in those decision making that we really want the A I to show its power and we're still at the early phase where we're dabbling with the visual side of things which is in and itself incredible already. But I do believe the future of A I goes beyond that where we will be allowing then building information modeling as we call it to allow A I to take control of those informations, large data sets, uh understand occupancy and how uh use requirements can be optimally adjusted for the final result while simultaneously providing the the end result visual so that you as a designer can both understand the A I is providing the most optimal solution, but also gives you the best design that you can identify visually. In the end, I think Keith touched on that point as well. Sometimes you, you see really nice visuals of the building but perhaps in reality, it's non deliverable. So, so I want to turn to one Y so you have you put some thought into this question like what can and what cannot not A I do, right? So tell us a bit more. Yeah, actually, I have a one slide to show that um since we are looking to the process of the design and sorry, let, yeah, let me just click. Yeah, to this one. So there's the two parts of the design. Firstly is that we build up our ideas just like Tim just did. And secondly, is what we just talked about that we need to develop the design into a proper deliverable, which need to do the um for instance, the regulation, compliance, budget, checking all those things. And this is another set of product which is the orange line. We have the X design cloud which is a composing by three elements, the data model and the rules. And this rule is not the so called rule based. It's a combining with three parts. One is the compliance, which is the regulation code from all regions. And the six is the logic. For instance, every design institute they have their own logic. And third is the pattern. The pattern is something A I can learn especially deep learning from huge set of data. So this is the rule and this rule is something representing the human brain, that is something we learn from our experience. And when we compose those three things together, then the result is not just images, it can be a 3D model. So our base of all those orange uh Softwares, we call it A I driven building information model on cloud in short ABC. And this ABC contains these three parts and can deliver to the end which meaning we can deliver to the fact to do the prefabrication deliverables. And this is the project we had with the uh the the fabricators before. OK, great. So um since we are now, I mean, facing the future already, so where can architects get some um in a real efficient way, get some basic A I training? Are there any crash course that you guys recommend. Yeah, I mean, it's like a deep blue ocean right now and we all new fishes trying to fish around it. And I think uh one of the main programs that I teach uh called parametric architecture. This platform, we have uh A I course as well where I teach there. So that's one option. Um like a lot of these platform, we also have many various like new architects that also uh work with these tools with A I. So everybody has their own observations and their own sort of um way of engaging with this technology for design. So I think it's interesting to, to have a look at the different perspectives from the different uh potential tutors you can find from the internet, right? When I know your app has a function, right? Yes. Perhaps spread the knowledge a bit. Yeah. Uh Also back to the slides, please. Um There is the one we call the Creation Plaza. Yeah, just this one, I can play it, I think. Yeah. And this is a show in the Christian Plaza. Every designer or users on this platform can share what they have created with the A I and you can see what the problem is and what's the base images. So you can learn from it. And furthermore, we can um allow the users to allow others to do iteration based on their generation, which means they can iterate further based on all those results by their by meaning with their prompts, their base images or with their combination of all algorithm models. So when you, you said something um in the in the recent article, you said current A I is only narrow intelligence, it can only complete specific tasks under the guidance of humans. So explain it a little bit more. What, what narrow intelligence is and what's next for. Uh look at. Is it going to widen the knowledge in in general that the A I technology is developing from narrow A I to A G I artificial generate uh general in general. Yeah, that is the path for the base technology. And also for the architecture design is we are facing a similar thing because now uh our, you can see this one, our Softwares are doing specific staging uh like design or engineering or calculation. But we already saw there's a possibility that using launch language model like chatbots, we can link all those different apps together to create a new workflow I mean from today on and this new workflow can allow the people to like communicate much better with the A I. So the narrow A I is no longer narrow each other and the people need to manage each of the Softwares. But here using the large model to link them. But in the future A I will just use all those background uh algorithms within the A G I. So this is the way how it's developing. But today, we can already use the A chatbot to using those apps as the background to do the whole procedure. Right? So Tim, once we've load, uploaded all the repetitive work to A I, where can um architects better use their, I mean, the save time and energy on what kind of projects? Uh For example, for you, um, are there any new territories for you to uh sort of make a foray into once you have? I mean, the A I as a genius assistant? Yeah, I mean, you know the fact that you can just reduce your work flow by a lot your lot of your workload because uh especially if you're working with complex geometries, you're doing a lot of calculations already when I was working as a Hadid, for example, I was in the advanced computational design team and we often use various algorithms to speed up our process. And as you can imagine with the typical geometry from that office, every corner, every piece is very distinct and requires a lot of these calculation optimizations. So in a way we already have uh uh you know, we very much engaged with advanced computation in order to reduce that work flow even before A I. So I think as A I comes onto the scene, it is actually doing similar things in reducing our our workload in terms of that. Now, with what we can do with our free time, with our freed up, I mean, we are ultimately architects and designers, right. That's the value that we can share to society. It's not that we can draw the floor plan better than other people, but actually, we can solve problems in a creative way and develop a design that can also fill aesthetic principles and design os that humanity and society would like to see. So I think those are fundamental human values. And because that idea this, I think that we're not threatened by the idea of A I. But actually we're hoping that it comes to do the jobs that kind of is more repetitive as you say, so that we can focus on what is truly important as a human to have input in. So in, in in summary, you get to work on more fun projects, right? More exciting and fun projects and leave all the heavy work fun is a way to put it. Yeah, definitely you want to be doing stuff that you feel artistically engaged and that's why we're in this career in the first place, you know. So I have a question for both of you. So there's a debate on what A I can do. One school of thought is a, I can democratize the knowledge base and even skill set. So ordinary people can become designers. But the other school of thought is that A I will widen the gap between the gifted and the average. So what's your take on this? Oh OK. I never thought about it in that sense. But I think um yeah, of course, if a gap exists between the people that have the skill sets to engage in these technologies to make better society, then they should fulfill the role to do. So I think um in that sense, yes, architects should be the ones that are trying to build a better world. Of, of course, they have to have those assets. And uh we're talking about democratization on a broader spectrum. Yes, always the more, I guess you could say the people who are visually uh has visual acuities and more talented in that sense will be the ones that prefer engaging in design. And what we do with democratizing it is you, you allow A I to be accessible to people all over the world. Nowadays, you no longer need such a heavy computer. And uh a big, for example, in order to render a high quality image. Now, someone from any part of the developing world can also partake in this process, very much uh wide leaning possibilities of what we can achieve as a, you know, collective humanity. But of course, every country, every place, they're always gonna be the people that will have inclinations to do these things in an artistic way when you, so your app democratizes a lot of these images and the knowledges in in in this very specialized area. So do you think we all can be become architects and designers using your app. Yeah, if it's for your own house, I think it is. But for the general public, like general public building, we also need the architects to really um create the ideas and also to thinking about how to balance the relationship between the general public and the client and the users. This is something, yeah, I think as a personal designer or like ordinary people want to be a designer, they cannot consider this part, right? You still need uh grasp and the knowledge of the industry. Yeah. Right. Thank you so much. I think it's a wonderful conversation. Thank you so much. Thank you.