Great to see you Back Gavin! and fantastic insight into the potentials of AI. Agree with everything and hopefully some interesting opportunities for beginners to move into more computational and programming roles with AI assistance, but as you mentioned programmers (experienced with language knowledge) are very much needed since AI isn't a magic pill.
Great insight, you should definitely look more into these tools with proceding videos. The more I've looked into the less weary I have been overall about AI systems in the AEC industry. They had a similar automation scare a decade and a half ago that didn't really pan out then either. In earnest though I am hopeful that these technologies will facilitate more linearity in the design process and make it faster and easier to coordinate between engineer/designer/developer. In that way I don't think we are far off a return of the industry to a very much design/build model, with the full AEC suite self contained in one roof with the hyper specialization sort of being done away with. In the far future, that may change, but for now I am very interested to see the direction of things.
I believe smart AI models will be made ( trained) for specific areas by professionals in all sectors; planners, QS's, project managers, engineers, lawyers, builders, code specialists, energy specialist, HVAC etc etc which will accrete to make a new way of creating, visualising, documenting and managing turnkey projects. I am certain that this is being done as we speak...Revit and Autodesk will be looking to acquire them as soon as they come to market.
I agree specialized AI is the real opportunity. I doubt autodesk will lead this space however, more likely to see a tech incumbent like microsoft or google take a swipe at that space.
@@AussieBIMGuruI agree, autodesk will not be part of the future rollout of a sophisticated AI solution = it will eat their bread and butter. Or they (autodesk/Adobe/archicad) have too much to lose by backing AI in the short term to keep their eyes on the long. It will be a big data operator working with very smart programers and specialists, and then self learning AI working with design professionals.
Hopefully it goes a lot further! I do worry a lot of AI apps are built off the hype cycle then don't go much further, but hoping to be proven wrong this time.
That kitchen example was pretty impressive. I guess AI design tools could become interesting for project developers who develop very basic and specific types of projects with very limited or specific design constraints, like self storage parks or other similar types of projects. We do a lot of structural engineering for such projects and a lot of them are very similar, and also occasionally do the building engineering for them as well. These sort of projects don’t require much architectural design, it’s mostly creating a site layout for precast storage units with some parking spaces around them. The zoning requirements are quite basic too. This might be an application where AI could have some value to help optimisme / iterate design options within a given set of constraints such as the number of parking spaces required etc. However I wouldn’t trust an AI to making actual structural engineering reports, safety is kind of important (look at the self driving car examples, sometimes they suddenly the hit breaks abruptly on the motorway which could lead to serious accidents). I don’t think it would ever replace a human, but it could become a useful assistive tool that should be used with careful consideration from a legal point of view.
There is certainly room for informed automation (maybe AI, maybe not?) when assessing structural designs at early stage. You could argue this is basic computation though unless the data has deep learning applied to it over time. I agree that trusting AI is a risk as ultimately the developers would never take responsibility over decisions, just that they informed someone to ultimately make one (so the decider is to blame).
Was looking for conceptual distributions for apartments in midjourney, also putting some promtps for floor plans, maybe this year it could be programmed.
Good topic I’ll listen more on my way to work tomorrow morning Think AI in architecture is on more and more My wife works for Kope they do some smart things for general contractors to optimise projects before they go to live with arch practices
Would be interesting to hear your thoughts about autodesks program Spacemaker, with utilizes AI in a way you did´nt really go through. Thanks for great content.
I've had my eye on spacemaker for a while, and so far it has yet to really blow me off my feet. It has some strong competitors such as Testfit, Archistar and Giraffe, all with deeper skin in the US and Australian markets. Spacemaker seems to be struggling to get outside the European market which is leading to a product misfit with Auodesk's US focus in my opinion, but hopefully they'll continue to grow and push. The environmental analysis and massing are OK, but nothing most firms haven't already been doing with Grasshopper/Ladybug for over a decade now. I can't say I was aware of any deep AI being used on S[acemaker beyond insight and trend tracking to give more relevant outcomes, which is probably a logic/data based process if so - and not AI in my opinion. Happy to be corrected (couldn't see any clear AI marketing on their website). The biggest challenge for me with Spacemaker coming from an architectural background is it's too simple and high level to go far beyond initial concepts and test fits of the site, and needs more intervenability like TestFit.
Great presentation Gavin. See all this AI emerging into the industry, a question always come to mind, will these tools distort the perseption of time each task requires.I know it will considerably increase the prodcutivity of many arch. offices, but i am a bit afraid that clients and managers, with time, will become a bit spoiled when it comes to how fast a task/project is completed. What do you think?
I agree that if we just boil things down to time and productivity gain we will miss the mark with AI. Our industry has an obsession with time and money, and needs to consider their value add more sensibly. BIM is a good example of how we save some time in areas, but lose it in others - AI will be the same. More time reviewing, adjusting and correcting outcomes, just like with BIM and people who can't use the tools we spend lots of time still doing redline markups and clash detection.
Thanks for showing little bit of how everything works. Haven’t had chance to play with image ai - nice to see in action. As for chatgtp, I found it borderline creepy good at trouble shooting it’s own code too. Let’s say it’s code gives you error, you can just copy and paste the error back to chatgtp - it will apologize 🤣 and then attempt again. Many times it gets stays stuck in a “wrong answer loop”, however, I’ve had plenty of times it got an answer right after a few tries of being wrong. That blew my mind.
I found the same, generally it does a better job of working with code versus queries - probably because it can match phrases more easily versus trying to process human language to code so it can do more with the same time period it takes for both when it's code to code.
Wow, that first picture of kitchen with Veras looks good and quite true to original. Your writing must have been golden. I was surprised because some of the words I'm not even that familiar with here in U.S. like polished joinery (rather than cabinetry) or word "benchtops." Good idea I guess to be fairly specific e.g. saying "close to viewer." Waterfall edge was very specific. Your description wasn't long but seemed to say the right things. Don't think the ChatGPT had anything advantage over what you wrote. I would like to see a higher "style" slider; heard that makes it more photorealistic. Yes you can't "tweak" areas as you say. I am using Enscape and want to see how this might be used with or along side those videos and renderings or not. Waiting for SketchUp Veras coming very soon. I like your additional examples, too.
I found fairly mixed results with AI and image refinement. Generally smaller interior shots end up closer to the original, and I suspect there's more images for relatable architectural interiors than exteriors for the model to draw upon. ChatGPT really only took my words and made them more AI friendly, but I agree it didn't really add much to the outcome in this case. Higher style generally deviates too far from the base shot for my liking - I'm worried AI will just become another tech that the juniors have to pilot for the non-tech higher ups, with even less control if we give the program too much control over the outcome. I'm siting on this sort of tech personally as I want to see senior/non-tech engage with it meaningfully before I'm willing to promote it to a company.
It's neat that AI can do this now, however I fail to see how this direction will be a "good" thing for anyone but maybe shareholders in tech and AEC companies. The list of potential negative consequences is long, complicated and should alarm every human. It includes thing like; the loss of life meaning found in human competency. The essence of Architecture is making human value judgements to tell the great human story. AI is an Alien Intelligence, it is not human. So yeah the "Gate keeping" is justified.
For the most part I agree. Since making this video it seems people are stumbling their way into this field with little regard for moral/ethical implications, and there is still so much hype around it. I think the technology needs to be explored and considered, but responsibly. Business minded people will always get the wrong idea with AI as all they want to do is ultimately make more and pay people less.
Is there any opportunities in Australia for BIM engineer ( U know it better ) as i am planning to complete my Master's from Australia ( intake is Sept 2023)
Yes, however it is a highly competitive and challenging market, especially for overseas applicants. Being onshore will help your chances but you will need to apply to as many opportunites as you can, and should try to attend and network at all possible BIM events. If you're studying on east coast consider looking into the state BIM group (SydBIM, MelBIM, BrisBIM etc.).
From what I have gathered in discussions with other people more experienced in prompt to image models, it appears to be utilizing stable diffusion. How much of the app is custom vs implementing an AI library is a valid query and one I also hope to see transparency around in future from developers using them.
@@AussieBIMGuru my exact thoughts. Several attempts to know whats happening in the background in very high level on linkedin but ignored. Thanks for the video! For the meantime we will have to deal with the AI spam until the industry stabilizes!
@@melsawy93 indeed we shall. Good on you for requesting more transparency, keep up the pressure - the only way we will see AI succeed is through honesty and openness.
haha I have been calling it mlWashing, everyone is claiming AI and machine learnng and then when you push them on wether or not its using machine learning or deep learning you find that its an algorythm after all. The exception maybe is Veras which does seem to be a true AI, although I suspect its basically just a reskin
Actually UA-cam recommend this video early this day,but I‘m quite busy today,also I’m not that interested.When I first know the word AI,it's just the difficulty of computer games,how smart the computer or so-calle bot would act. In recent years,in China,many articles and news are talking about “5G.AI lot blockchain,AR VR or wearalbe equipment,Auto Pilot,Big-Database ....”They say technology are changing people's dayily life.But it turns out a lot company are just boasting,they just show a lot advertisemen but never invent anything or have product.Once they got the money they flew.so people are tired and see these company as joke,they just use new ideas to cheat investment. Thus a lot ordinary people like me in China are not that interested even chatGPT is popular these days.But I’d like to see new things come out as long as they are real technologies not some ideas to let rubbish company cheat for money.
I agree entirely - many companies are not genuine about AI (I discuss that in this video as well, I'm not 100% on the side of where it's all going). It will however make some serious changes in how we work once we get past the hype cycle - keep an eye on it.
Great to see you Back Gavin! and fantastic insight into the potentials of AI. Agree with everything and hopefully some interesting opportunities for beginners to move into more computational and programming roles with AI assistance, but as you mentioned programmers (experienced with language knowledge) are very much needed since AI isn't a magic pill.
Yes if people wait for the magic pill I fear by that point programs will just replace the people who could potentially program.
Hey Gavin, Awesome to see you back on youtube!
Cheers Daniel!
Great insight, you should definitely look more into these tools with proceding videos.
The more I've looked into the less weary I have been overall about AI systems in the AEC industry. They had a similar automation scare a decade and a half ago that didn't really pan out then either. In earnest though I am hopeful that these technologies will facilitate more linearity in the design process and make it faster and easier to coordinate between engineer/designer/developer. In that way I don't think we are far off a return of the industry to a very much design/build model, with the full AEC suite self contained in one roof with the hyper specialization sort of being done away with. In the far future, that may change, but for now I am very interested to see the direction of things.
Will do!
Yes for the average non AAA firm nothing will change (just the fees going down as always). Fortune favors the bold.
I believe smart AI models will be made ( trained) for specific areas by professionals in all sectors; planners, QS's, project managers, engineers, lawyers, builders, code specialists, energy specialist, HVAC etc etc which will accrete to make a new way of creating, visualising, documenting and managing turnkey projects. I am certain that this is being done as we speak...Revit and Autodesk will be looking to acquire them as soon as they come to market.
I agree specialized AI is the real opportunity. I doubt autodesk will lead this space however, more likely to see a tech incumbent like microsoft or google take a swipe at that space.
@@AussieBIMGuruI agree, autodesk will not be part of the future rollout of a sophisticated AI solution = it will eat their bread and butter. Or they (autodesk/Adobe/archicad) have too much to lose by backing AI in the short term to keep their eyes on the long. It will be a big data operator working with very smart programers and specialists, and then self learning AI working with design professionals.
I'm using veras yo get suggestions and some quick sketch for clients, really excited too to see where this will go
Hopefully it goes a lot further! I do worry a lot of AI apps are built off the hype cycle then don't go much further, but hoping to be proven wrong this time.
That kitchen example was pretty impressive. I guess AI design tools could become interesting for project developers who develop very basic and specific types of projects with very limited or specific design constraints, like self storage parks or other similar types of projects. We do a lot of structural engineering for such projects and a lot of them are very similar, and also occasionally do the building engineering for them as well. These sort of projects don’t require much architectural design, it’s mostly creating a site layout for precast storage units with some parking spaces around them. The zoning requirements are quite basic too. This might be an application where AI could have some value to help optimisme / iterate design options within a given set of constraints such as the number of parking spaces required etc. However I wouldn’t trust an AI to making actual structural engineering reports, safety is kind of important (look at the self driving car examples, sometimes they suddenly the hit breaks abruptly on the motorway which could lead to serious accidents). I don’t think it would ever replace a human, but it could become a useful assistive tool that should be used with careful consideration from a legal point of view.
There is certainly room for informed automation (maybe AI, maybe not?) when assessing structural designs at early stage. You could argue this is basic computation though unless the data has deep learning applied to it over time. I agree that trusting AI is a risk as ultimately the developers would never take responsibility over decisions, just that they informed someone to ultimately make one (so the decider is to blame).
Great video Gavin,
Thanks!
Was looking for conceptual distributions for apartments in midjourney, also putting some promtps for floor plans, maybe this year it could be programmed.
A tool will no doubt emerge at some point, but currently the prompt results are too broad for such detailed/specific outcomes based on my testing.
Good topic I’ll listen more on my way to work tomorrow morning
Think AI in architecture is on more and more
My wife works for Kope they do some smart things for general contractors to optimise projects before they go to live with arch practices
Yes AI is certainly changing a lot of industries where insights can assist early on with decision making in more depth.
Would be interesting to hear your thoughts about autodesks program Spacemaker, with utilizes AI in a way you did´nt really go through. Thanks for great content.
I've had my eye on spacemaker for a while, and so far it has yet to really blow me off my feet. It has some strong competitors such as Testfit, Archistar and Giraffe, all with deeper skin in the US and Australian markets.
Spacemaker seems to be struggling to get outside the European market which is leading to a product misfit with Auodesk's US focus in my opinion, but hopefully they'll continue to grow and push.
The environmental analysis and massing are OK, but nothing most firms haven't already been doing with Grasshopper/Ladybug for over a decade now. I can't say I was aware of any deep AI being used on S[acemaker beyond insight and trend tracking to give more relevant outcomes, which is probably a logic/data based process if so - and not AI in my opinion. Happy to be corrected (couldn't see any clear AI marketing on their website).
The biggest challenge for me with Spacemaker coming from an architectural background is it's too simple and high level to go far beyond initial concepts and test fits of the site, and needs more intervenability like TestFit.
Great presentation Gavin. See all this AI emerging into the industry, a question always come to mind, will these tools distort the perseption of time each task requires.I know it will considerably increase the prodcutivity of many arch. offices, but i am a bit afraid that clients and managers, with time, will become a bit spoiled when it comes to how fast a task/project is completed. What do you think?
I agree that if we just boil things down to time and productivity gain we will miss the mark with AI. Our industry has an obsession with time and money, and needs to consider their value add more sensibly. BIM is a good example of how we save some time in areas, but lose it in others - AI will be the same. More time reviewing, adjusting and correcting outcomes, just like with BIM and people who can't use the tools we spend lots of time still doing redline markups and clash detection.
Thanks for showing little bit of how everything works. Haven’t had chance to play with image ai - nice to see in action.
As for chatgtp, I found it borderline creepy good at trouble shooting it’s own code too. Let’s say it’s code gives you error, you can just copy and paste the error back to chatgtp - it will apologize 🤣 and then attempt again. Many times it gets stays stuck in a “wrong answer loop”, however, I’ve had plenty of times it got an answer right after a few tries of being wrong. That blew my mind.
I found the same, generally it does a better job of working with code versus queries - probably because it can match phrases more easily versus trying to process human language to code so it can do more with the same time period it takes for both when it's code to code.
I tried several external renders. Giving up in the end , and go back using lumion ....
Yes it's not ready to replace detailed rendering workflows as of yet.
Wow, that first picture of kitchen with Veras looks good and quite true to original. Your writing must have been golden. I was surprised because some of the words I'm not even that familiar with here in U.S. like polished joinery (rather than cabinetry) or word "benchtops." Good idea I guess to be fairly specific e.g. saying "close to viewer." Waterfall edge was very specific. Your description wasn't long but seemed to say the right things. Don't think the ChatGPT had anything advantage over what you wrote. I would like to see a higher "style" slider; heard that makes it more photorealistic. Yes you can't "tweak" areas as you say. I am using Enscape and want to see how this might be used with or along side those videos and renderings or not. Waiting for SketchUp Veras coming very soon. I like your additional examples, too.
I found fairly mixed results with AI and image refinement. Generally smaller interior shots end up closer to the original, and I suspect there's more images for relatable architectural interiors than exteriors for the model to draw upon. ChatGPT really only took my words and made them more AI friendly, but I agree it didn't really add much to the outcome in this case. Higher style generally deviates too far from the base shot for my liking - I'm worried AI will just become another tech that the juniors have to pilot for the non-tech higher ups, with even less control if we give the program too much control over the outcome. I'm siting on this sort of tech personally as I want to see senior/non-tech engage with it meaningfully before I'm willing to promote it to a company.
It's neat that AI can do this now, however I fail to see how this direction will be a "good" thing for anyone but maybe shareholders in tech and AEC companies. The list of potential negative consequences is long, complicated and should alarm every human. It includes thing like; the loss of life meaning found in human competency.
The essence of Architecture is making human value judgements to tell the great human story. AI is an Alien Intelligence, it is not human. So yeah the "Gate keeping" is justified.
For the most part I agree. Since making this video it seems people are stumbling their way into this field with little regard for moral/ethical implications, and there is still so much hype around it. I think the technology needs to be explored and considered, but responsibly. Business minded people will always get the wrong idea with AI as all they want to do is ultimately make more and pay people less.
Is there any opportunities in Australia for BIM engineer ( U know it better ) as i am planning to complete my Master's from Australia ( intake is Sept 2023)
Yes, however it is a highly competitive and challenging market, especially for overseas applicants. Being onshore will help your chances but you will need to apply to as many opportunites as you can, and should try to attend and network at all possible BIM events. If you're studying on east coast consider looking into the state BIM group (SydBIM, MelBIM, BrisBIM etc.).
What model does Veras run on? I have asked this question several times with no response.
From what I have gathered in discussions with other people more experienced in prompt to image models, it appears to be utilizing stable diffusion. How much of the app is custom vs implementing an AI library is a valid query and one I also hope to see transparency around in future from developers using them.
@@AussieBIMGuru my exact thoughts. Several attempts to know whats happening in the background in very high level on linkedin but ignored. Thanks for the video! For the meantime we will have to deal with the AI spam until the industry stabilizes!
@@melsawy93 indeed we shall. Good on you for requesting more transparency, keep up the pressure - the only way we will see AI succeed is through honesty and openness.
haha I have been calling it mlWashing, everyone is claiming AI and machine learnng and then when you push them on wether or not its using machine learning or deep learning you find that its an algorythm after all. The exception maybe is Veras which does seem to be a true AI, although I suspect its basically just a reskin
There's certainly a lot of wash out there - the key is to try it out and decide for ourselves I think.
Actually UA-cam recommend this video early this day,but I‘m quite busy today,also I’m not that interested.When I first know the word AI,it's just the difficulty of computer games,how smart the computer or so-calle bot would act.
In recent years,in China,many articles and news are talking about “5G.AI lot blockchain,AR VR or wearalbe equipment,Auto Pilot,Big-Database ....”They say technology are changing people's dayily life.But it turns out a lot company are just boasting,they just show a lot advertisemen but never invent anything or have product.Once they got the money they flew.so people are tired and see these company as joke,they just use new ideas to cheat investment.
Thus a lot ordinary people like me in China are not that interested even chatGPT is popular these days.But I’d like to see new things come out as long as they are real technologies not some ideas to let rubbish company cheat for money.
I agree entirely - many companies are not genuine about AI (I discuss that in this video as well, I'm not 100% on the side of where it's all going). It will however make some serious changes in how we work once we get past the hype cycle - keep an eye on it.