#47 AI Architecture Revolution with LookX.ai founder Wanyu He

what's up guys and welcome back to another episode here on the architect Network today we're going to be jumping back into the crazy world of AI and we're going to be doing this with wanu her who is the co-founder of Lux Ai and also the company xcool both these companies are super interesting you've probably heard of Lux AI more recently it is the world's first architecture trained diffusion model and you can go and check it out I've got the link uh it underneath the video where you can go and uh add your prompt and it's trained in architecture stuff so you can pick out your favorite architect whether the the image it will create is like a model or a render or something like that it's pretty incredible tool and also look xai was kind of uh created in excool and xool is a company that she founded in 2016 very interesting company that is somewhat of a tech company for Architects uh where they make AI software for Architects to use and they kind of work in a number of different ways sometimes they kind of consult with Architects to help them use some of these tools sometimes they collaborate with them on competitions they've even done competitions using these tools as a proof of concept and actually won some of the competitions which is super interesting uh so we're going to be delving into all of that with Wu and also talking of course about the future I think W is really a pioneer of AI in our industry she's been utilizing it since 2016 way before all of this mid Journey era and image diffusion models and so uh we're going to be talking about uh AI before that era the the era that we're in right now and also what's in store for us for the future as always though if you like this video give us a like and a subscribe and share this with your friends and colleagues so that we can share this knowledge around uh you can always follow us on Instagram and Linkedin where we post upcoming talks videos and events that we have so stay stay tuned for all of that but without any further Ado let's jump into it and have a chat with one you all right that's it we're live one you thank you so much for coming on to the podcast today I'm very excited to have a chat with you hello Lee it's my pleasure to be here it's really honored and uh looking forward to talk to you today great uh so I thought it was again AI has been a big topic over the last uh few well every podcast right now um but I wanted to jump back onto this top topic because there's so many things going on and uh obviously you guys have released a super interesting tool that's that's uh uh changing uh our industry a lot but having looked into your background and stuff I was really amazed at all the kind of things your you're very much a Pioneer in the space so uh we won't just be talking about Lux uh AI will'll be talking about your whole background leading up to that but just to give the audience a little bit of a background in case they're kind of uh a little bit new to I've given you an introduction in the intro but I always do these little bullet points uh so it's kind of interesting your back looking into your background you've you're very much uh come from a an architecture background of of real world architecture building stuff right you've spent eight years at uh Oma and I think we were just discussing you worked on the shenzen stock exchange yeah which must been super interesting you've spent a little bit time in Academia you've you've taught in Hong Kong you and also you did your doctorate at in Florida FIU yeah uh where you studied Ai and space architecture which is a very interesting combination uh and then you went on to found uh uh ex cool which uh I understand stands for ex cool house but I don't know if I if that's true uh and of course look at AI more well somewhat recently um so yeah I think you you've got this really interesting background of uh actually coming from working in practice to being more in this Tech space uh which I think is super interesting uh so I think an interesting point to maybe start off with you is that transition from um you know working as an architect in in om right and I think uh you studied at at Del and where you I think you were kind of studying uh some stuff in computation and and the technology side of things where do you feel that this transition came from the architecture world to more like Ai and of of course space architecture and well as well where did this transition and and stuff come were you using computation and Technology a lot in in the days or was it something you kind of pursued uh after uh actually before om just directly to answer your question yeah because I was a studying in bog Institute now it's part of the tood and when we uh start to choose our Direction I there was two paths one is a more theoretic and another is a more like combined to the technology part and I choose the the technology the reason is I always be fond of the new tech the future while I was a k so this is always what I reaching to joining and then I study with all those things but after I join om found wow it's different than I thought I mean Architects work is totally different it's not that much uh Advanced let's say it's a quite quite traditional yeah yeah and even even the the like a really the the top off office like om still a lot of the I mean bunch of the um Talent people we sometimes was wasting our talents there because we're not just dealing with the Innovation we're also dealing with a lot of dumb things change quite a lot and the time and time again repeating some works and that that part made me feel yeah confused in the beginning and later on feel no I cannot stand on it anymore I should do something to change it yeah that was a long procedure yeah that's super interesting that you brought that up because I think it's true there's there's um even some of the the top films in the world there's still a lot of antiquated slow things we we repeat even myself I'm I'm always talking to Architects like hey stop copying that around we can just that quickly and and stuff like that it always seems somewhat of an uphill battle um but yeah that that's interesting that that you raised that point um okay so that yeah that's super interesting so when you're working in om were were you also trying to utilize tools like um like computation and things like that or uh or you at that point were even experimenting with very early um AI software or anything like that unfortunately no you know I was trying to especially my first years in rot then H I only found a few chance to apply what I have learned from barlock that is the algorithm design and uh there was a only two times I think I can use it first time was a uh we do some kind of faad and then they wanted some like variations of faad looks more interesting this is the one case another is a uh try to allocate different Master plans but that option wasn't be adopt most the time om doing things with blue forms so we have to call Blue forms a lot and then draw Cad and then building Rhino that's what we do and uh yeah in Rhino sometimes you can use some scripts to reduce the amount of work but still yeah that is not the the the most uh yeah heavy load part of work yeah yeah and to be fair actually I I because you often see of these different conferences and stuff like that now you now you mentioned I almost never see people from o om I do get the sense they have a slightly more traditional design process I should try and find their uh a few people from the OM Tech Team and and uh reach out to them for to come and show us what they're doing but uh but yeah I can imagine that so that that's interesting and then uh of course you then went on to found ex cool um and so maybe maybe explain to the audience what is excool and and how did that come about because I've been listening to a few of your podcast and and it's been super interesting you're you're kind of a tech company but then you've also been doing these architecture projects and competitions and stuff like that so maybe talk to us about how did xool start and and what is it yeah uh actually it start with the idea that we want to use a new tech to change this industry at least the for ourselves in the beginning and this is something I've learned from om that's why I didn't go to like a un studio in in Netherlands or zaha in London because I I think the most valuable thing of om is the thinking the thoughts right and that is also the key of architecture why was like just the the the the yber Freshman into barlock Herman herburger as the thing said What is the or he asked us what is architecture in your mind and people said a lot of things like space light circulation functions and he said architecture is about thinking and that is the something really new to me uh yeah because I I thought architecture is space and experience but he think he say something totally different and also I found in rims uh Theory it's also about thinking it's about uh like all the arguments all your positioning all your scenarios those kind of thing I think is the most important part of the architecture of the design and that value is not be I mean uh revealed sometimes because a lot of people spend too much time on the form or on the functions and that is something I think yeah I think that is something we should release The Architects from yeah that's the starting point of EXC we want to relieve ourselves uh first first of all of course myself I don't want to draw this PL again and again like to calculate what is the basement yeah that's that's insane especially when you become a like a senior architect people expecting you to understand all those details of the architect so you have to look into the details with the regulations everything also like talk to all those Engineers from MEP or Etc but that is not the the most the valuable part and uh I was always into the valueable part of a understanding the circumstances of the project and what's the uh demands or the requirements of the client as well as the society and we found our own position and we can find the The crucial point and our argument and we can set up our scenarios and then is our solution and design that's the last thing the first part of all those procedure is the most valuable part and I not today I still appreciate that part and this is something I also try to talk to my students about that is the key of architecture so uh back to X cool is um uh it like you said X coool is also like extremely cool also like uh yeah want to want to carry on something out from Rim Yeah and that is the thinking part so we think the valuable thing is the thinking is the operational research and decision making and that is the most valuable part but rest we can estimate maybe some parts of the work is interesting and it's valuable but the value is not just you draw it but you can see it and you make decision I made some kind of a uh uh example of like Oma you have a at least the three levels of Architects first is the the the uh junior and or interns and the freshman and second is the senior and Architects and the top of those Partners but people do different things in different level the Juniors always doing models even the inter coming in do the model all the way all the time and there is something I don't think they fight is fun but they also uh they also F some fun of it isn't when they making different options and those options you will find is a it's a kind of a typology actually it's the strategy to do it those senior Architects or the project Architects will lead a group of young guys to do okay you do this direction and you do that direction and that those guys do another Direction and each of the small team or each person can create five or uh yeah whatsoever three or five options with the models and you Gathering Together find wow 50 of it and this is the what they do but the project active what they do they say okay this category we don't like it this category doesn't fulfill the requirements this category is not uh super for the content so we do deductions we cut out all those option we make decision in out fromont those 50 uh different possibilities and then we got a two options out saying okay Ram you choose which one which one you like yeah this is a the final decision Final Call he will make however he will not look into the form he will ask some really interesting question about it for instance there was a project I did with him in uh quo it was a competition qualo Financial Center and there was two options one is a uh more like a circular shape and uh yeah some something pops out from it it's a really big scale another one is I did with the the algorithm to generate different uh volumes according to the the side condition the circulation Etc finally I didn't ask anything about the form he was asking who is the party there what's their agenda and uh what is their vision of and we say okay this part is a new and they want to uh Union or reunion all the different religions different races and that's their yeah a way to to uh propaganda themselves and then say okay we select the one can show or represent this unionship then we choose this or he choose this circular typology so this is how he make decision and because he got his position and he got his perspective of things so this is the most valuable part and uh and this is what we do as a tech company we provide options to The Architects and help them to make right decisions so this is a yeah the philosophy behind how we establish this company yeah that's super interesting I think um you're obviously uh you know I work at Big which is a what they call a a baby REM right so our yeah yeah our design process very much follows uh that similar process and I've been having a lot of conversations recently about this of course this whole AI image generation thing which is is mindblowing uh but also I'm kind of within the office I'm I'm kind of like seeing where are we going to use this because on one hand the magic is exactly what you said the thinking the the actual research I'm always like uh in admiration of the teams how much research they put into and it's it's super interesting to see like the the questions they have sounds like a very similar process to what you had at Oma whereas I look at other Architects that are maybe more style based like uh maybe like a zaha or or a Gary or something where they seem to deliberate over the form and I could see how image generation can be completely used in that kind of office and scenario because it's they are kind of starting with the image in a way or or the image is the the kind of somewhat the end product whereas in what you were describing is a much more process driven and and the form is really uh a reflection of the performance of the building or or or whatever is dictating so so I think there is a really interesting comparison there of like this AI era we're in and the type of architect uh your you know you are or the practice you work for is um which I think is is really interesting I was watching some of your your videos of this this kind of tool that you have where you're able to analyze and then look at all the information that you have and then come up with with many different options and you even use this right in um or or you have used this or other arches have used this in competitions and was it at this point you started to use this technology to enter competitions with the your own software and tools that you had right yeah we actually team up with Architects we barely do it our by our own we try to empower certain Architects and they they are embracing us so we also uh support them with all our capabilities some are more mature software some are like algorithm still some are something we test in an app so this is a really fun collaboration uh we know this Architects have their certain like um aspects or their their way of thinking and that is somehow we can really uh maximize the tech part and they can also maximize their thinking part and this is a good combination we tried it several times and we really got a really good results out on those yeah yeah I um I'll have to put in the video you did with uh with NY and on the digital Futures right where you were explaining that you had a competition and I think you you you'd won it or you were one of the top top uh picks and then they found out that you were a tech company not necessarily like an architecture company they were like oh this is are you allowed to be in this competition or something yeah yeah yeah and it's it seemed like a really pioneering way where you're kind of like testing your your technology and ideas but in a real world scenario um and I thought that was a a really interesting uh example of of your using this and the interesting thing about all this was this was not uh recently like within the last year year or so when we've been living in this I guess mid Journey era right this was kind of uh a few years ago so you were on the pioneering end of of using these tools and I guess back then you you would maybe fall under the umbrella of generative design right this idea that you you feed a bunch of information or parameters and it generates a bunch of options uh but really you using uh different forms of AI would would that be correct yes we start in 2016 and why we start in that year because we inspired by uh the the alpha go yeah yeah that that is really the reason because uh like a even five years ahead then 2016 was 2011 we start work with I mean I start with work with my partner and we won a competition uh by his name and did this agon design generative design at that moment we thinking maybe we can do something like differently I can quit from and you can create from your New York office we can do something like interesting but at the moment 2011 we found this impossible the reason why is there's no new technology there's no new algorithms it's all like a traditional uh like all those logical agdon and uh machine learning expert system Etc and this is one thing AG not that compelling compared to today and second we don't have data there's no data coming out because uh at the moment just the starting of a Mobile smartphones and and third thing there's no computational power we did the algorithm uh to generate our stuff but it like take out three days to get results it's insane so yeah at the moment it's a yeah not possible but 2016 everything become makes sense because there's a deep ning coming out conary new networks it can mimicing stuff and to generate stuff not from algorithm not from logic it was out from data and this is give us a really good example because doing the Go game it's really like you allocate the the for instance the houses on the the plot or you allocate your rooms in the floor plan or you allocate all the components in a room so we found out wow this is something we can really start to work with so we started in 2016 and start to use all those deep learning algorithms first the CNN then G generative adversary new networks and nowadays it's a more like a diffusion model and some other agreements as well so it's a different path yes because we found out the traditional way of a generative or parametric design doesn't really work yeah that that already back to 2011 so uh in in the day one of exu We call xu. we have our um our website and everything be set Based On A Cloud Based on data driven based on agent driven yeah wow and so now ex cool um is that something that do you have a lot of Architects using that is that something that someone listening can go and test out right now or how does it work now is this a tool available for people um to use uh today excl the software is a mostly like supporting our So-Cal members yeah those members are The Architects we would like to work with and to support them with all our capabilities so we team up for competitions or even some commissions so it's a different way it's not a subscription that you can find online it's more like you're almost like an AI consultant in a way would like uh we also provide them softwares no they can also use it but sometimes we are Consultants yeah we provide them some solutions it's more like that when we work with Architects but the uh then looks AI is different it's something we can like for the public use everyone can join me to use it so it's a different mindset because the previous one we were still like testing and it's more uh in detail it's not the 2D Graphics it's a more 2D uh geometry and 3D geometries so it's more precise into the real projects yeah yeah that's that's really interesting so yeah as you said you've now like lookx I guess gave gave birth so to speak uh to lookx AI which was the kind of image diffusion generation thing if if uh people from the audience haven't tested it I'll put the link underneath the video but you are you you are or were one of the first ba architecture based uh AI diffusion models right where people could go and uh you could put in a prompt you could pick a different architect or a style like a a physical model versus a render and things like that and it was a the first kind of architecture trained AI right so yeah was that how did that come come about was it something you developed obviously in EX in uh in ex cool and then you decided this is this could be a really amazing tool for everyone and kind of turned it into a product so to speak yeah yeah actually uh we did the researches about image generation back to 2018 and in 2019 we also uh issued a paper about it we call this building does not exist it was using Gans to generate images and I think I also shared some work uh with with new nich's Forum like digital future online and that moment people say wow this is really cool can you go it a bit further but with against we can only do like a uh smaller size of the images and it's not precise it's not like diffusion model you can really uh be diluted to the r small particles and make it really nice but gen uh generative adversary neuron networks is still workable with some certain pathes but the uh the effect or the outcome is not as perfect as today so at the moment we think um no this is not yet to be a product it can be some of our capability we can offer to the architect but it's not a public platform yet so it was like a long time study already and then we have a Today l x and all those data we had some of those were generated out from the previous models and some we collected from our collaborators Etc yeah and uh it's of course kind of taken the architecture World by by storm I think uh you know we're in this I guess it's the the AI diffusion era right now where it's becoming I guess the diffusion side has made AI much more mainstream for Architects it's it's very tangible for them to type something in and they get what looks like a a render um how do you see either your tool look X XI or um this diffusion model where where do we kind of sit within the um within AI in the industry because this diffusion model for example from from my perspective it's mind-blowing like I said I I'm still like looking at many different people using it of course you got the people on social media that are kind of posting things every day and blowing up on social media you've obviously we've been testing it in the practice I've been testing it in my own like ways and I've really been trying to see where it fits into the design process because like you said uh that render it or like like that image is just an output of this thinking process um but then you see other people that are kind of using AI as the kind of inspiration or they they start with the image and work backwards which is also quite interesting but what's your perspective of the era that we're in now particularly with this AI uh image generation era in in our industry um I think um in different uh countries or different areas Architects they they took it differently if this country is still under massive development I think most Architects will be forced to do fast work so they do the image generation then they work backwards to make some geometry look like it I think this is a yeah this is the background they are facing I mean the the real world they are facing uh we we cannot judge that of course if you can step back and think oh this is not the the right way to do it then you can really follow the the the way I think is a more logical to Think Through what's the real issue here and what's the problem we would like to solve and what's our positioning to do so and what's the argument could be then you can make decisions out from those uh yeah nice images because the imag is just the images everyone can do image there's those images doesn't make you any like a uh competitive power out it no because the AI actually give the rights to everyone can do the same thing and what is the value of you then that is the value of Architects I mean this is this mainstream you cannot stop it it's coming so yeah we have to think like what is the the real role of Architects we're not just making images making models because eventually of course the generative a will make 3D models we already did in some certain typologies but then when that come so what will you do we're loading jobs no I don't think that is the the the solution and that is the end of Architects I think that is the chance Architects can really rethink what is our role what is our value I still uh completely agree with the the the Dutch uh philosophy that the Architects bring thinking to the project can really throw thinking to answer the question of the society the the question of the environment Etc so when we facing all those options we should really take back and think what is the question and what I want to do here it's not just selecting some nice images yeah yeah exactly no I think it's a really nice uh sentence like uh architect architecture is thinking Architects are thinking and uh you know I get asked this a lot like is AI going to take over jobs and all this kind of stuff and for sure it will disrupt things but uh like you I I think it it will allow us to concentrate more on where we add value which is this thinking process and reduce the time that we do coloring in stuff on Photoshop or like drawing plans line by line and stuff like that and and free up that side of things and I think that's where it's the most interesting is like I think the era we're moving into this AI collaborative era uh and I actually think you the work that you've been doing uh with with uh with cool and and all this kind of stuff is is really pioneering particularly those competitions where you're kind of really really utilizing these tools your your Architects collaborating with AI to create new and interesting proposals uh and I I really think that's the future um and the other thing I I heard from from that talk was the team you you um that were working on this product was actually very small right um yeah and I think AI gives you the gives also gives you that power that you could potentially see more or a more diverse range of of firms like smaller firms that are able to compete and tackle on large products because they're aided by some form of AI whether that's you know diffusion models or or whatever it is um so I think that's what's excites me about the future I'm not so scared that uh you know the robots are coming so to speak and take over our jobs yeah but uh but Architects do need to pay attention to us and and not ignore it because I think we we have a natural tendency to be like dismissive of this thing particularly with this AI image generation where you're like ah it's just an image just you know next but um it's really important to to kind of stay on on track with these developments yeah given that you have a bit of insight into this and and you know you've been working with this for almost a decade at this point how far do do you get do you have a sense of how far away we are from a 3D models which I I know there's various different uh platforms that already you type in stuff and you get a 3D model but I guess specifically for us as otic we're we're almost talking about a Bim model like do you do you see this as something in the near future because everything moving so quickly or will it take time to be able to generate you know models that you can actually use to build things uh it's a good question uh because today everyone is looking into 3D uh geometry Generation Um not just the architecture uh industry also like a film or uh game especially however they don't really need precise 3D model as we do so they are more like generating those um image envelopes that will be uh comparatively easier let's say of course when want to be more like detailed and looks the uh real of course need some time there's some research predict need three to five years to make it as mature as today's image however this is this is something not that relate to what we do as Architects because archit don't want just envelope we need geometry to be precise and that is something uh what x cool was doing in uh from the beginning on because we thought there is a new form of uh the architectural data is be uh required in this new era um actually there was a standard that ISO create called the smart standard uh it's from level one to level five level one just um uh drawings and sketches and level two is scanned U those drawings level three is called a machine readable document like Cad and you can edit it you can save it and copy it and then level four is a machine readable content it's not just the the content itself but it's a readable means that it has some data in it so that's a uh to our industry is more like a beam but in the future it's a new typology it's called a machine interpretable content that machine can self-generate self- learning self iterate and even machine can like U communicate to each other through this kind of new format and in that way it's some it requires some new type of a data which is not exist today and we we were trying to do so and we we start this one we call the uh AI driven building information model on cloud and that is uh something new because it's a old cloud-based and it's not a rivet beam it's a beam that according to the IFC standards but it's in 3D geometry and with this new type type of form we can have the machine generate 3D objects uh with a precise geometry the vectors and all the information the most important thing is it's not just the content model as the cad and beam all have and it also contains the data beam also also have it but it also contains the most important thing is we call the rules but this rules is not just the regulations it's also the rules it's like the law you abstract from large amount of data machine learned from all all those big data so those are things where the capability or part of the capability of a human Architects to understand the regulation to understand the code to understand the the the rules for instance for a company they have certain rules so those things we can really um embedded into this system and the result of this is that machine can start to generate things out from it we call it's a more like a data driven mode which is different from the traditional way we always do as a graphic driven mode what is graphic driven mode it's like a I have a thought I will draw it out and this is the graphic and other things needed in this industry CH for instance what's the cost what's the regulation checking you have to measure it out time and time again this is the loop we always lost the in between because when we talk to the client his imagination our imagination different and when we do things our imagination and our hands can make something it's different there second Gap and third Gap is when the design be made out is it fit for the regulation the requirement the code have no idea have to test it and this is third Gap and fourth Gap is when everything's done you go to the cost estimation wow it's over budget sorry you have to do budget have to Value engineering then you change it again and then the in the end when you go to the site the Ste you have some problem you have to change it so all those gaps what's the problem with it is because we were starting from graphics and the graphics doesn't contain any information any rules any reasons behind it so we have to change manually we see the data those are numbers the numbers incorrect then we change the graphic because these two things is not associative to each other it has to be like a you you change out from those uh uh like simulation or consumption or the estimations you have to change it so this thing eventually is the problem of our current data format because the data format is not advanced yet even beam is advancer than the cad because it contains the information however beam requires too much effort to do it you know in China there's no Architects want to do beam they always join in cat then they threw this uh the cat drawing to some like a render company they build the beam for them it's not Beam for the design it's the beam to just as a kind of a submission document so this is this is not the correct way because it's a notot that uh reducing the work but adding the workload for Architects so you don't want to use it and we think the good way is that okay machine can help you to generate this and machine can help you to change if you change the data the model will change if you change the model the data will change but if you change the regulation everything can be changed so this is these three objects or not objects three elements are more really uh Associated to together and in that way we have realiz the one particular uh field which is mic modular integrated construction from the M planning from the beginning towards uh you can make it in a factory and assemblying in on site it's it's in entire chain we made it uh I mean with AI it possible so it's not just the one one uh one one part of it as the design or as the cost or as the connector Factory it's the entire chain we tested a project I think several days ago or a couple of weeks ago Guardians also issued a report about it and took a image um yet it was not the nice image because it's a mic it's a modeler of course it look like a camp it's just the Extrusion of the building but with this mod we can do a lot of things yeah it's not the fault of the machine but this is the fault of architect decision making you know so the tool just Ste tools but the tool can help us to really yeah to to release our some our time to really focus on the thinking part but if the the time be released and the people still too lazy to think to make decisions of course you get the bad results yeah yeah that's uh I'm still processing that super that's sorry too many no yeah it's super so you've got like um to to kind of summarize it a little bit you've got like almost you got super dumb a super dumb model which is CAD or let's say a rhino model it's just geometry that we use it's completely graphical we literally just use it as a way to create renders and drawings and then we've got this the middleman which is Bim which is uh it's still graphical but there's information inside of it what you're describing is I don't know if there's a word word for it yet but like we'll call it AI we call it ABC yeah we call ABC in short okay AB ABC what does ABC stand for the AI driven beam and Cloud ah AI driven beam and Cloud okay a so then we got ABC where it's essentially you know geometry with information inside of it but it's intelligent like uh when you're saying the rules inside of it of course you're talking about uh regulations and all this kind of stuff but also that the wall knows that uh a wind it's good to place a window here because it's sou facing and it's this high off the ground because that's how tall a human is and and all this kind of stuff would that be you know what you mean by part of the rule sets yes um it's a it's a very similar to how we uh use the AI to generate the the they the master plan our first product of xool is to generate master plan with the different uh Tower typologies and this is a uh the way that we are first of trying to combining deep learning agreement like CNN with the the expert system the expert system is more like a uh regulation based or rule based you have to to uh make those constraints to fulfill the requirements but the flexibility or the variations you need to be generated by the CN to get a lot of possibility which can uh be fill to those those requirements through the experience of the previous projects so machine learned from previous projects and mimicking it so in this way it's very similar to the details we do with the um for instance in a room uh the reason why we choose mic because it's all uh like dfma designed for manufacturing and assembling all those components got standards and uh yeah it you know it's from the factory and all those components assembling you need some rules to assemble together it's very similar to the master planning so in this way I always think Architects can deal with any scales it's the same machine can deal with all the scales as well in previously I mean the the the last generation or even two generations ago The Architects doing the masterer planning the building the interior even the furniture everything because it's a similar in terms of a thinking in terms of logic and then today the machine can do the SAR thing yeah that is the How We Do It from the large scale towards the small scale because the room be composed by the components and the room can compose into a unit and unit can compose to a like a floor plan of the building and the floor plan can compose the building itself and the building itself can also compose into a larger uh block so this is a like um top down and bottom up can merge together through through the machine because previously people or Architects we always can only do certain things in certain step for instance we have to do a very rough sketch in the beginning then we do details for instance I draw a poetry of a people I have to do around the shape and then I do brow eyes and nose mouth but then next step I do more details and detail and detail again but machine don't don't do this what machine do is like this one time everything done that's because it's a processing parallel in large scale middle scale and small scale this is how machine do things yeah so so in that process then what is the role of the architect are you are you suggesting that like it's coming up with many solutions and you're kind of pointing it in a direction a little bit like uh how you were describing how REM uh cool house would be the that one at the end that looks at these two uh options ask questions and then say maybe go forwards with that or create some other variations of that is that what you kind of see how does the how does the architect uh interject in this kind of process yeah I think so and um uh there's a certain moments Architects can interfere with the machine like we we chopped into steps firstly your inputs that's the most important thing you need to give order to the machine the machine just machine it cannot uh yeah it's not a super intelligence yet so you have to do tell the machine what to do it's just like you ask the right question that is the most important thing so in the beginning Architects can really guide the machine to make generate content and then they generate a lot contents then you need to make the decision which direction to go and there's a lot of steps to make those decisions and eventually you can get what you want with the help of the machine and we also in the same time when we design software there's some little um details of it we have H always two M of this kind of 3D geometry generation softwares uh first way is the you input your parameters your demands and the Machine gener outcomes to you they give you 10 and maybe there's already 10,000 behind it yeah this is one way you make decision you put input and you make DEC that's it this is the one way and the second way is the machine is a more be guided by the architect for instance there's a plot we can draw this area I want highrise this area I want certain type of a uh like a low rise or this area I want some vas so this is a more like a architect still do the graphics but do the zon Lings no more like the the guidance or you can do some a to it and yeah the machine can do whatever you want it to do so there are two ways one is a yeah you you can rely on the machine the machine generate stuff to you and you make decision and another is you work with it really in details yeah that's that's really it's mindblowing yeah maybe I can send you some videos then later you can add on it's a more easy to expand that then I think that' be great because then people will be able to actually see it but that's it's really incredible um what do you feel like do you have any concerns for the future of Architects are you excited about the future of Architects or should young Architects be concerned about this rise of AI I've even had some some uh young students come be like is there any point in studying architecture because they see this mid Journey thing and I'm like yeah of course you know this is this is just an image at the end of the day but uh what's your thoughts on the future of of architecture with AI what do you have obviously your pro uh it coming but like should there be some concerns with with Architects like would would we all be kind of controlled by whatever this algorithm is behind whatever AI model that that they have um yeah what's what's what's your view on the future of like Architects and and architecture practice I think first all uh if Architects are more positive to embrace the new tech they will find that there's a lot of possibilities in the future and there was some young kids I mean uh just in the in the university the students they came to me say what we should do after we graduate should we change our when we still in the University there any job for us in the future I said of course when you go to any City you just step into those alleys there's a lot of things you can do I mean they're lacking of design everywhere because there's a lot of problem not be detected not be realized not be fixed those things are the jobs of Architects come on the Big World big cities even small cities small streets there's a lot of things to do so first of all we have a demand of course always and secondly the AI is steel tools it's just like cars we are not reasoning with the cars come on we reason with people and we drive cars so when some people drive cars you still want to walk no problem that's your style that's your choice but people driving cars will make their speed much faster than the people who walk this is just different way of working however sorry you want to say something no no go on go on yeah yeah yeah it's okay just interrupting me anytime yeah so I think the the car is uh is similar as today as the AI and in the future perhaps uh yeah we are not uh uh driving cars anymore there's autopilot or like a the car can drive itself yeah but we need to give directions where to go right so the most important question again is where to go is not a how to drive yeah that's a very that's a very nice analogy actually I do like that I was I was going to challenge you and be like does AI I actually push back when people say AI is just a tool because I I feel like AI some of the AI stuff is crossing the line between a tool where it's almost like a collaborator um and and no longer a tool like you know a a grasshopper script for example that takes a facade and breaks it breaks into panels and you know it's reacting because of the Sun or something it's a very linear process it's not really adding anything to the conversation it's just essentially saving you a lot of time modeling right yeah but uh some of the stuff some of the examples I've seen from from what you guys do or even the image Generation stuff it's it could do what an intern for example could do or a junior architect um it's adding ideas and it's bringing ideas to the table and so at that point does it cross the line between a a tool to a collaborator in in some sense do you think there's there's something in seeing AI less as a tool and more as a collaboration uh the reason why I still think it's a tool because uh I think there's no um thinking in it so the previous uh thinking procedure we were talking about is not exist in the machine yet I see except AGI so still today it can do certain things it's more like a narrow AI in certain task it can do perfect task ask however it cannot do the chain of thinking and that is the key we consider this is our collaborator it's a contribute not just a one point it's the entire thing yeah okay that that's a that's interesting so until we got AGI it's still that's the that's the Sci-Fi going to yeah the Sci-Fi side okay super interesting well I know we're almost up to now so but I uh it's it's super been super interested talking to you maybe a few like final questions of like yeah let's say there's a young architect at University listening to this podcast um I think there's two things that are kind of interesting one is what you're doing and you're one of the few people I see doing actual AI because there's a lot of people now that are like hey I'm an AI architect and they're kind of Masters at Mid Journey or using chat GPT and also every plugin right now is suddenly overnight uh AI powered because they've kind of they use the chat GPT API to like you know fill out some information or or something it's really hard to find like actual AI if if you know what I mean so as a young architect how what advice would you give if if they were really interested in pursuing like jumping into AI for AR or equipping themselves with with uh these AI tools for Architects what what would you advise for them to kind of look into or study or or go and find this knowledge right now yeah I think uh first thing uh in the university the kids is to really understand what is the the the role the mission of Architects and then you will find what tool you need yeah even though today there's some tool not smart yet but you needed to do something just work with it it doesn't matter but you need to understand the most important thing is who you are and then you use tools whatever tools you will be confident even it's AI tool you you won't feel oh it's taking my job so uh I think uh first thing is to understand the role and secondly yes if you want to uh like say what is the real AI tool to help us then study some like Tech uh basic uh theories let's say we actually we wrote a book about the we call it the the guide book to Architects about Ai and yeah but sorry was that was in Chinese maybe we can translate someday yeah that was the yeah that was the book for our internal team because in our team there are coders and Architects some don't understand each other and we hope them to understand each other so we start to writing some like a guidance inscriptions and what is the background what is the theory behind it Etc so we wrote this thing then um someday we published it I think in 2020 uh Neo claim is the first book about AI Architects but yeah that was the book we have and with it I think people will understand oh AI is not new there's also machine learning before and AI is a trend there's a lot of models coming in and going on and they will understand where to go in terms of the technology and then they will have the confidence to choose what type of a technology they want to adopt what type of a software or tool tools they want to adopt so yeah try to learn more about it then you will see it's not that difficult to to find out which one is your favorite yeah yeah I almost wonder do do you think that's also the future uh composition of architecture practice like it'll be Architects and uh you know technology people uh together or I would almost I almost wonder like should you start uh architecture bachelors and then uh used do um CS like computer science as a Masters or something like that and you come up as like a half cyborg half architect like uh background do you do you do you yeah what do you see the the future of practice is it more of a mix or these hybrid kind of characters I think each of the direction will have its own value uh however if you want to do uh architect sorry if you want to do architecture in a more technical way you don't have to study computer science be honest because computer science is really a large Team every source of a vower required industry with one or two people it's very difficult to devel something out of course with some others resources some open source yes of course you can use to do something but if you want to create something with a small team is a very difficult in especially today so uh if you want to understand the logic behind of course you always need to take the course of a AI or computer science I heard there's like a primary school they have a AI course already so check it from your yeah high school or so it doesn't matter yeah if you like computer science you can just go for it it doesn't matter yeah and I always think cross discipline is the future of every uh those disciplines it's not just architecture everything because all the industry I mean our University was uh designed for industry Society all the majors are fit to each of the industry but in the future won't be like that a lot of things mix together and all the old industry or traditional industry will be reformed by the new text so it's good to combine all the time yeah yeah cool and and my final question I I'm asking everyone at the moment is is there any technology or thing on the horizon that is really exciting you and of course for you you're not allowed to answer anything AI is there is there something that's coming up that's you see maybe plugging into what you're doing for example this game engine World unreal and all this kind of stuff like I almost find those two worlds are colliding right is is there something like that on the horizon that is really exciting you um in terms of like the a space I guess you mentioned the uh um MMC and and stuff like that but uh yeah what's what's exciting you outside of AI uh I I think there's two things one the virual world or like more into the visual like a Vision Pro I think that is something uh can really mix our our virtual experience with the reality and this is something that the Architects might really dealing with in the future we might not build that much in the physical world anymore perhaps but still there's the issue need to resolve of course but we also have a bunch of people doing virual design in the yeah in that space mix the reality with the virtual uh experience so this is something I think is very interesting and the other is the the space technology I always so into it I think we are we are heading to this uh interplanet uh species yeah we need to go to the moon to the Mars to everywhere else yeah I see yeah we didn't even get to touch on Space architecture yeah unfortunately have to get you back for a special uh space AI architect uh discussion because there there's some really interesting uh combinations about how AI can be used to expand uh traveling to you know yeah Mars and Beyond which is super exciting so yeah we'll get you back for AI space talk because it's also my uh my new favorite topic as well so um great my pleasure those more to discuss yeah well thank you so much for joining I think this has been one of my one of my most mind-blowing talks uh I've had on the podcast so um thank you so much for coming I I really think you are one of the pioneers of this AI uh movement for Architects uh and I'm definitely going to keep a close eye on what you guys are doing and I'll put as many links as I can find in the link in the video below so people can go and check out excore X look Ai and and all the other kind of stuff you're doing so and hopefully let me know if that book is ever translated into English uh because it sounds like a good one or maybe I have to really quickly learn Chinese but we can translate it with the help of a CH Beauty exactly yeah we'll just we'll just do that well yeah thank you so much and uh thanks everyone for tuning in hopefully W you we'll have you on uh on the podcast almost uh in the future maybe we we need do it regularly to catch up on all the amazing stuff you're doing um and we'll be back in a in a few weeks with our next podcast I think we're going to stay on the topic of AI so uh we'll keep this discussion going but uh thank you once again Wu and we'll see you guys very soon thank you oie thank you everyone nice to talk to you all right Cheers Cheers

Channel: ArchiTech Network

Views: 6,131

Keywords: Architecture, Architect, Technology, ArchiTech, ATN, modeling, computation, bim, visualization, graphic design, immersive, augmented reality, virtual reality, game engines, parametric architecture, grasshopper3d, parametric, metaverse, blockchain, nft, clubhouse, video game, Space Architecture, 3D Printing, AI, AIarchitecture, AIInArchitecture, Artificial Intelligence, artificial intelligence, lookxai

Id: vZ5bpXOHQcQ

Length: 66min 19sec (3979 seconds)

Published: Sun Nov 12 2023