AI is NOT Your Mid-Century Design Friend

You may be already using AI to get input on many areas of your life. And you may feel like the convenience outweigh its many negatives. Maybe. But here on this podcast where we talk about making right choices for your home, let me assure you, you will not get good advice or even peace of mind about your options by asking AI.

In Today's Episode You'll Hear:

• Why I’m so irked with AI.
• How AI will lead you astray on your remodeling journey. 
• Where to find better answers for your home and your life. 

Get the full show notes with all the trimmings at https://www.midmod-midwest.com/2304

Like and subscribe at Apple | Spotify | YouTube. Want us to create your mid-century master plan? Apply here! Or get my course,  Ready to Remodel.

Transcript

00:00
Please, my dear mid mod remodeler, do not ask AI for remodeling advice, especially for a mid-century home. Okay, I've had this rant building up for quite a while now, and I am generously and pretty vociferously in the anti AI camp now. You may be already using AI search or chat GPT to get input in other parts of your life, if so, you may find that the benefits of convenience outweigh its many negatives, but here on this podcast where we talk about making right choices for your home, let me assure you, you will not get good advice or even peace of mind about your options from asking AI. 

00:37
Hey there. Welcome back to mid mod remodel. This is the show about updating MCM homes, helping you match a mid-century home to your modern life. I'm your host. Della Hansmann, architect and mid-century ranch enthusiast. You're listening to Episode 2304. 

00:53
Okay, before we get into the fact that I think asking chat GPT for information about specifically, a mid-century remodel has about as much likelihood of being right as a stopped clock. I honestly, I think one in 12. Odds are maybe generous. I will tell you that the person I think you ought to ask for advice about your mid-century remodel is me, although you can ask a host of other people, including Joe from Reddit 10 years ago, to anybody inside the mid mod remodel Facebook group lovely folks all to your friends on the block who also live in mid-century houses and might like some of their features or be able to tell you the history of the houses they've lived in through their generations. 

01:32
If you want help, directly from me, from mid mounted West in planning your remodel, though, particularly if you are thinking about doing some work in this calendar year, in 2026 now is the right time to get in touch. Now is the right time to start a conversation. So I basically just want to start this episode by saying I'm not scary. You don't have to know everything about your plans and have everything gamed out before you reach out and start a conversation with me about your house before you schedule a call to find out if a master plan is even what you might want. 

02:06
The best way to figure out if a master plan approach is right for you, for your household, for your house, is to talk to me about it. I've got an easy form on my website. If you go to mid mod dash midwest.com/services, and click Apply to work with us, I will ask you a couple of quick and easy questions about your house when it was built, what kind of shape it's in, what are your general hopes and thoughts about it? What do you like about it? That's mid-century. That will let me know what kind of conversation we're going to have. And then I'll reach back out to schedule a time for us to talk about your house, what's possible about it, what we might design for it, and how that design process could go. 

02:43
The next master plan project we get on the books will probably not be getting its design options back until March, maybe April, at this point. So now's the time to get the ball rolling. It all starts with an easy conversation. I hope I hear from you sometime soon. Get that process started or just go find the transcript of this episode in the show notes at midmod-midwest.com/ 2304. 

03:10
I'm going to paraphrase a meme I saw recently that said, Hey, y'all don't ask ChatGPT Make friends with a nerd. Ask us we use a lot less water. In fact, you'll probably need to remind us to drink water, and this is true. A ChatGPT does use a horrifying amount of water, and B, someone you can ask who has a personal level of expertise about it, whether that's me, a friend of yours, or Joe who posted on Reddit 10 years ago, you're going to get more reliable, more reasonable and more human centered data out of a person than you are out of chat. 

03:49
I'm going to tell a slightly silly on myself anecdote, which, speaking as someone who obsessively researches things, a deep nerd, I hold a lot of information in my head, but occasionally I want to go back to a source, and mostly I can put my fingers on it, and sometimes I cannot. An example of when I can't find a source, I once referred to is I read a book, a popular nonfiction nonsense thing, a list of, I think, 100 practical items to have in your home that might have surprised you. Maybe the only 100 things in your you need in your home. I can't tell you the title of this book, because the whole point is that I can't remember the title of this book, but it was the first time I had encountered the concepts of either an electric kettle or electric blankets as ways to heat up water or yourself in a more energy efficient manner. I thought the author was charming, and I would love to read that book again. 

04:41
Every now and then, when I'm bored and trying to stay off social media, I search for that book again, trying to put prompt search terms into Google that will give it back to me. But recently, I did this, and I before I could scan my eye down the. List of actual website hits that might or might not have been relevant, and immediately roll them in or out, using my own human judgment, the Google AI prompt came up. I put in a search term, something like book, colon, something like I thought the title was, and then I put in a comma and a couple more options of electric kettle, electric blanket, hoping that would prompt it, I actually thought, Ooh, this is new information. I could put in maybe I'll get an answer this time. 

05:22
But this time, but what I got was a hilariously or offensively aggressively wrong answer from Google's AI summary, which told me the book you're looking for, and then it rattled off a title that was nothing like what I had just asked it for as a book title in quotation marks, doesn't contain any information on electric blankets, blah, blah, blah, blah, blah. Now, okay, this is an obvious example of put in kind of garbage information, get kind of garbage out. And I, as the person who knew exactly what I was looking for, was able to immediately recognize that I was getting absolute gobbledygook from the Google AI summary. 

06:01
But if I had instead asked it a complicated question of, I don't know something I was trying to remember from high school and only vaguely get recalled physics or a moment in history that I was not familiar with, and I had gotten back a result from Google's AI that was just as confidently delivered, I wouldn't have had the knowledge base to determine that it was outrageously, ridiculously, not even the question I had answered asked wrong. And this is really part of what I find most pernicious about AI, and what I don't think is ever going to get any better, because it is designed by its creators to give us confident answers whether or not it has accurate information. 

06:47
This for me as a as a professional who gets asked questions all the time, I've done whole episodes on the nature of answering questions. I love a Q and A session, and when I get asked a question about residential remodeling, about mid-century history, about making the right choice for you, for your family, for your house, I try to give every person an answer, but the answer that I give, and the tone in which I give it is always colored by my level of confidence in the specifics of that information. I'm a person who likes to give people options as well. I'm a big believer that there's not usually one right answer, there are multiple correct potential answers, or there are answers that could be correct based on circumstances or information that has not been included in the question. 

07:30
And so this worry that I have, this fear that I have, that we are all falling into the trap of just asking the internet for information, we have a tendency to go Google for whatever we want. I do it myself. I just recently was comparing social security numbers with my little sister for reasons of family estate planning, and realized that even though I'm older, her number is larger than mine. We were both weirded out by that. So I instantly turned to Google and said, Why would a social security number have a different cause. Actually, the thing I asked was, we were born in two different states. Does the state you were born in affect your social security number? 

08:07
And I got a bunch of website hits which indicated that yes, the state you're born in seems to affect your social security number. But this is a matter of passing curiosity being instantly answered based on everybody having a really short attention span these days, and which I am also so guilty of, but I think that trap of I'll just Google, that leads us into a trust of information we find on the internet, and the fact that the internet was originally created by a whole bunch of nerds who were just filling it up with technical specs and personal reflections can take us into a place where we are way too willing to take the advice of the internet. 

08:48
Because I've been irked by AI for a while. I'm always kind of on the lookout for more information to corroborate my opinion. Again, human nature. Do with that what you will as you interpret this data into your brain. But the Washington Post recently ran a top of the page headline piece on what AI can do versus what humans can do. I think the title was something like, Can ai do your job? And the very first example they gave, actually, was, can it do the job of an architect or the job of a designer? And they tested this in a bunch of different professions by asking a person to do a task and asking AI to do a task. Some of the other examples were things like, put together a data dashboard based on a bunch of information. I think they used World Happiness scores, they fed it some data, and they asked it to put it together to create a video showing a product based on rendered information to create a simple game.

09:51
In each case, the AI alternative didn't do a particularly good job. I was particularly caught up on their very first example was asking it to create. A digital floor plan based on a hand drawn floor plan. And it was literally it was just supposed to recreate the hand drawn sketch as a digital floor plan. The human version, very tidy, clearly done in some kind of CAD program, neat and annotated with dimensions. The chat version, just a bunch of rooms labeled and with a bizarre relocation of spaces, including a bathroom surrounded on three sides by corridors. 

10:31
Okay, so clearly, they were looking for an example that would be obviously and hilariously wrong. But the point is, AI is not going to create our floor plans for us anytime soon. Beyond that, though, AI is generally not a good way to start to continue to get your referent information, even if you're just asking it a simple question, like, what kind of trim is appropriate for a mid-century house? I'm going to talk a little bit more about all the different ways that that question could go wrong. But as I was sort of tracking news articles about AI and how it works for my plan to deliver this podcast episode before I could even get my notes in order, I popped up another major headline news item about AI, which is that just a few weeks ago, Google announced to advertisers that they have a new kind of ad that can be bought in AI mode, Google's chatbot style search is now going to have the ability to include ads with all of its answers. 

11:40
Look the era of free and ad free AI use is apparently already over, and when you compare that with the relatively long lifespan, I mean, I guess in our lifespan, it's not that long, but in the lifespan of the internet, of relatively low ad social media, this is coming at us like a freight train. Just imagine, in five minutes, basically, we're going to be getting completely ad driven results from our AI question searches. And sure, we get ads in Google searches as well. 

12:16
There's definitely always sponsored media right up at the top, but it's easy to identify and scroll by, and the very nature of the way that answers are delivered in this sort of human sounding conversational chat in AI search responses means that it's going to be much harder to detect ad information in AI searches. I want to get forward into why AI is not a great idea for design thinking, for you, specifically, for mid-century advice, generally. But I wanted to just first hit a couple of bullet point topic elements of why AI is bad for the world in general. 

12:56
The short, short version of that would be that it is burning up, or rather, more specifically, it is drinking up our natural resources, the water use and energy use of data centers that are required to support all of these large language models and AI search modes are truly appalling and not we're not regulating them fast enough. We're not doing anything about it fast enough asking each of us to individually assume responsibility for that. It's a little bit like making everyone feel guilty for not recycling properly, when what we actually should be doing is regulating corporate interests. But if you need a reason to avoid AI searches, it's there also, in the same vein, basically every result of an AI type search a large language model is intellectual property theft. 

13:44
Where does your AI get its information? Okay? Caveat, I'm not the expert on this here. Do a little bit more searching, reading, get some books about this to learn more. But we should all understand that when you ask a large language model like chat, GPT for information, it's not actually scanning the internet for you. It's instead drawing from an existing database of information that it contains because it was quote trained unquote on it. So chat, GPT four doesn't have access to data more recent than its most recent, so called training conclusion. 

14:18
Where they get that training information is a past snapshot of the internet. It's Wikipedia, it's books, and by the way, there's a lot of debate out there about the fairness and the legality of AI companies taking a big net to scoop up the internet, all of the intellectual property and copyrighted material, published books, art, journalism, countless examples of human ingenuity have been scraped for profit, and there is and already will be more profit. 

14:46
So two things of that to note, asking chat isn't as up to date as just asking the Internet if you're looking for somebody's most recent idea, and it's crappy to benefit from this mass scale intellectual property theft to do it. Sam Altman. And the other AI tech execs are well aware of the dubious legality of their process, and they basically have two answers for this. The first one is that they needed all that copyrighted data to build their models, so they had to take it so they did to which I say, I'm sorry, what? And two is basically, if you don't like it, come and sue me about it. If you've got the time and the resources for that. 

15:22
Lawsuits about all of this are in the pipeline right now, but we'll see how quickly that can be accomplished and what kind of guardrails can be put around this rapidly unregulated, expanding industry. The third reason, though, just as a moral consideration, to avoid hopscotching, getting someone to do a little bit of this work for you is professional development, if you care which. It's not everybody's individual responsibility, but as an architect, as a person who benefited from generational professional experiences spending my first years out of grad school doing a lot of low level scut work, picking up red lines for my boss, designing little bathroom fixtures to proper specifications, doing a bunch of basically boring intellectual grunt work that did never seem to benefit me much, but I knew I had to get it done for a paycheck. The sort of thing that we can now have AI resources at least start to replace. 

16:26
I find this kind of horrifying that we are looking for ways to circumvent, not cheap labor, but beginner labor. It's one of the things that I total digression, but one of the things I'm proud of in running the business of Mid Mod Midwest is that in our process of developing sketches for all of our clients, I do some sketch work myself, and I also have two design assistants who are both current or recent architecture design students, and they are getting the opportunity to flex their sketching muscles, To practice over and over again, designing little details, putting together options that I give them feedback on and respond to them about they are learning step by step, one mistake at a time, one clever idea responded to at a time, how to put together the mechanics of a house. 

17:17
This is true industry by industry. It makes me sad when journalists aren't getting the entry points into their careers that they got from putting together pretty dumb little listicles for early internet websites that technicians in many professions just aren't getting the sort of baseline starter work that they need not just to get a foothold to become an intern who then becomes an assistant, who then becomes a designer, who then becomes an named architect, who then becomes a partner in the firm, but also to build up the knowledge base, the muscle memory of intellectual professional habits that go into the process of becoming someone who has design details at their fingertips. 

18:01
So that all of this is why it's bad for the world, but I want to talk specifically about how AI is not a successful way for you, as a homeowner of a mid-century home to get information about what is a good or bad idea for Your house. Because ultimately, if you are asking chat GBT or putting it into an AI search mode, what is the right kind of light fixture for my house? Or should I put on addition in the back?

18:31
The ultimate purpose of the question you're asking has a couple of key fundamental details that AI is never going to get right for our purposes here, anyone who's here listening to mid Monroe model. You're curious about mid-century quality and correctness, and this is something that any large language model or search AI function is extremely unlikely to get right. It doesn't have any ability to discern what is or is not good data, and there is a lot of bad data out there, as I've talked about many times before, the search term mid-century has been very polluted by its trendiness. 

19:10
In the last decade, it's been a profitable thing to label a modern product. This is a mid-century coffee table. Here is some mid-century art on society six the fact that oh so SO, SO SO, SO MANY magazine articles on mid-century, this and mid-century that which are not actually about historical correct things at all, but are just nice looking, cute designs that are perhaps even vaguely mid-century in character, but not in any way specific or Correct. Chat GPT does not know how to determine the difference between that and actual historically correct information, even if you use vintage in your search question. 

19:49
But beyond even that, the purpose of asking a design question is to determine what is important to you, and that's a question that. Than I can't answer. If you were to ask me, what is the right choice, I can provide you options. This is something I'm always interested in helping my clients, my students, people that I talk with about their mid-century homes to determine these are questions I am more likely to reflect to you. You're going to have to ask yourself, though, what's important to you? What do you like? Because again, when we ask the question of what is mid-century appropriate. 

20:22
Not only can chat not determine whether something is genuinely mid-century or just mid-century branded by a corporate marketing office in the last 10 years, it certainly can't determine, for example, where it falls on the level of mid-century appropriateness of updating, but using some qualities of mid-century design, the sort of x and y axis of your perfect mid-century moment and style, it's never going to be able to get those answers right. Getting good design always comes out of generating answers that come from asking the right questions. 

20:57
And AI just doesn't it depends on your prompts to develop anything, and we'll confidently give you garbage data, even if you've asked a question carefully, but certainly with our tendency to whip out the phone and just, I'm just going to Google that. I'm just going to google how the numbering system of social security numbers works. It's very easy to get bad data and then internalize it. I want to talk about the antidote to this, and the antidote to asking for computer generated images, putting materials into your kitchen so that you can judge whether they're right or not, is good old fashioned sketching and good old fashioned getting actual physical copies of something and containing them in the same space together to see if they work together. 

21:44
So I want to talk a little bit about what you can do instead. What I do instead in my design process, when I'm working to solve the potential problems in someone home a rather than confidently giving them an answer. I am always providing them with a range of possibilities, and I'm trying to render those possibilities, and I use that term maybe a little incorrectly here, I'm trying to present those possibilities in as non answery of a way as possible. I want to take a little side quest here into not the question of AI, but to the question of computer generated imagery, or computer graphics in design at all. I am not advocating we return to the day of literal blueprints. 

22:34
For one thing, the way that blueprints were done was full of toxic chemicals, and for another, the idea of hunching over a giant drafting table just gives me back issues even thinking about it. But the advent of modern CAD computer aided design is what CAD all caps stands for. Gave us has its pros and cons. It's certainly been a marvelous tool in terms of efficiency, it's allowed us to generate more accurate data about buildings and to create complicated drawing sets, edit them without starting from scratch, and easily disseminate them to everybody that might need that information at the click of a button. It's great, but tools are tools. 

23:22
The right tool for the job is really important. It's always easy enough to go ahead and use a tool if it makes you feel confident, efficient and happy, but I think it's also really important for us to be aware of the risks and influences of the tools that we use. With a hammer in your hand, everything looks like a nail, and we tend to use the tools of our design software in ways that can limit our scope. This is true even of architects and designers. 

23:49
By the way, just as I was getting out of grad school, and I'm going to go ahead and put a date stamp on myself by saying that was in the mid aughts, we were getting out of the era of AutoCAD and getting into the era of AutoCAD, Revit and other similar 3d modeling tools. These allow you to build a comprehensive model of the building, which isn't just lines, it's made up of materials. So if you make a building that the walls are concrete block, inside the model, there are a bunch of little concrete blocks, and it accounts for the fact that you shouldn't put a doorway in the middle of a block line, but you should put it where the block's edge is for more efficient construction. 

24:23
It allows you to generate real Windows rather than just a rectangle and get everything set up almost from the very beginning of the process in great detail. Once a building has been designed this way, you can then cut a floor plan horizontally or a section vertically, and any change you make to the model itself will be properly ramified out into all of the various drawing sets generated from it in an interlinked way. This is, I'm sure, a revelation for the designers and constructors of large new buildings, but it has its problems.

25:00
It has its sort of built in, easy to do moves. And I think that you can see the effects of, for example, the extrude tool, which just allows you to select an area, maybe a rectangle, maybe a loop, maybe a hollow square, and pull it forward or push it backward. As you drive down the street of your town, you might see recent construction, anything done in the last 10 years, and see the extruded shapes in the awning design and the facade design windows that seem to stick out for no particular reason, other than to be interesting, but don't actually seem like a particularly good construction design. 

25:38
We have this ability to use the tools at our disposal copy and paste something or push and pull it using computer generated ideas that limit or focus our creativity. This is the same reason, by the way, that I advise homeowners not to get hung up using a room planner software to plan their remodel. 

26:02
Or if you do use it, to use it with extreme caution, it's possible to see very clearly the size of a particularly actual product, or constant sized things, common size things, to move sofas around a room. For example, to group a sofa two chairs and a coffee table on a rug and rotate it to put art on a wall and see how that might look, quote, unquote, in a lightly rendered 3d space for people who don't have a good sense of reading a floor plan, this can be very useful for visualizing, but it's not a great way to design because it limits you to thinking about the rooms as rooms. 

26:41
Rather than questioning whether daylight can connect a space, whether a better connection could be created by moving a doorway, by shifting a wall, by opening up two separate spaces to each other, making more dramatic rearrangements to your footprint. At the end of the day, I would always rather see someone thinking about design in a chicken scratch floor plan, on the back of a napkin or on graph paper. 

27:06
Rather than have a client send me their export of some online design tool, it's more likely to be based in what's truly important to them, even if it's not proportionally accurate. And then my job can be to figure out the mandatory minimum spaces or proportions. Or for that person, they can start with a chicken scratch floor plan on the back of a napkin and then go into something that is more dimensionally accurate. Shift from napkin to graph paper to the computer in that order to make sure that the level of precision in your tool is appropriate for the level of decisiveness that you've got going on. 

27:40
This is actually also why, or at least a big part of why I and my team do not use a lot of computer modeling in our design process now, full transparency, we do actually use the program SketchUp, which many other designers actually do use, to put together buildings and generate even floor plans and section and elevation drawings and full drawing sets. That's not what we use it for, but you can use it for that. 

28:07
But unlike the CAD program Revit, which, for example, if you want to draw a wall, is going to ask you, what is that wall made of? Is it framed in two by fours? Does it have drywall on each side? What is the thickness of the drywall? It wants to know all of that information before you can draw a line. SketchUp just lets you make a sloppy square, but you can build in as much precision as you want, and it's not creating any inappropriate levels of accuracy for information that's not available. 

28:34
This is where we get into the question of, and I'm circling back around to the philosophical topic of, don't ask AI for advice on your remodel for your mid-century house. Remodeled houses are a whole other ball game. When you're dealing with an existing house, it is not built the way it was planned. I guarantee you, if you have access to your original blueprints, there is something there is some window that's out of place, there is some floor thickness that's not as drawn. 

29:01
There is going to be something that on the day in the field on site was not done the way it was planned to be done. Buildings are filled with quirky irregularities, if they have been remodeled in the past. They are filled with secrets and mysterious structural changes and odd references to the original design that show up in the 80s remodel mysteries that can only be solved with a great deal of speculation and never really known for sure, and when we are dealing with remodeling information, when we're putting together a floor plan, even if I'm the person that's on site taking information. 

29:35
But certainly if I'm getting floor plan information from my clients, I am not getting a computer aided design level of accuracy out of our FOIA plan, and we don't need it for creative schematic design planning, building in hypothetical accuracy that we don't actually know precision without accuracy. To be more specific, really irks me. I feel it's a dangerous thing when you are telling a computer that you're moving a wall by an eighth of an inch, when actually you don't know that down to the eighth of an inch, you are creating false information, rather than vague information, and this is one of the reasons why sketching is so effective for the kind of big picture creative thinking that we need to do to make all of the possibilities that are great happen in a design process. I use this myself. 

30:27
So as I said, we use SketchUp in our client design process. We use it to build a rough model of the floor plan of the house. It comes in most handy for making sure that we're not going to run a counter in front of a window and that spaces add up to the inch in our kitchens and our bathroom designs, we use it to set up easy perspective drawing so that we can more efficiently generate the helpful perspective images that we share with our clients. But we begin in sketch and we end in Sketch. 

30:56
Just this last week, I was struggling with the complicated details of three scheme variations to remodel the layout of an existing space with a kitchen den and laundry area for a master plan project. Now my team had put together a lovely model of the space based on some blueprint drawings that an architect had generated in the 80s when they last remodeled the house, and I was using it fairly regularly in my scheme thinking, because this house has a complicated massing with high, clear story windows and two different roof lines across the space. 

31:28
So I wanted to make sure I was considering the sort of 3d nature of it, not just the Floor Plan nature of it, the different ceiling heights, the different adjacent rooms, but ultimately, the more time I spent in the SketchUp model while I was still trying to create the big moves for the three kitchen schemes in the space, the more I got hung up, the more I was focusing on granular design possibilities. How would this countertop work exactly? 

31:53
Oh, was I really putting in a kick plate for the baseboard in each space I got hung up and distracted on unnecessary details, spending too much time chasing down little design rabbit holes, the answer in every time I got hung up to my mental gridlock was to come back to the sketch, to put my pen style as fairly broad and just outline the general idea of what was cool, what was interesting, and more specifically, what were three different from each other and distinctly useful designs for our client to consider, which was the most preserving of the existing footprint and layout and structure the least construction cost, and which would be the most dramatic design move that was possible without transforming the entire house from start to finish. 

32:35
And then what was a good compromise in between them? That answer came out of sketching by hand every time. Then, because this is a complicated project, it went back into the model for rough massing so we could easily do our proportions. And then those massing models allow us to cut some views and turn them back into sketches where we put in all of the detail, the detail that helps our clients visualize which option is going to work best for them. That process, by the way, is why we occasionally get clients who at the end of the process, after they've gotten their master plan document, which is full of beautiful sketches, sketched floor plans, sketched views, they'll sometimes come back and ask for the model or the computer files that are the outcome of the process. 

33:18
And it's a good moment for me to remind them that I told them at the start that we don't have computer model files to share, that rough SketchUp massing model isn't worth sharing because it's quick and dirty for our own purposes, and often the last details of design are being finalized at the sketch level in the sketch floor plan, and then ramifying, using our own human intelligence up into our views, rather than constantly going back and tweaking the model as a construction documentation record. That's not to say that computer models aren't useful most of the time. 

33:49
The reason our clients are asking for them is that they're about to work with a contractor who wants to work up a quick computer generated model of the whole project from which he, she or they will cut floor plans and also render up some views to show more of the built ins and cabinetry and whatnot, in perspective to our clients. Now, this is an interesting thing, because at a recent Office Hours call, I had a past Master Plan come with an example. She'd just been talking to her contractor, a design build firm that is doing lovely work. They were using top of the line computer graphics to render what the kitchen would look like with a few modifications she'd made from our original schematic master plan and all of the built ins suggested by their cabinet shop. 

34:34
She was worried, though, because the images that she had gotten from them suddenly didn't seem mid-century enough. They looked too hard modern. They looked too contemporary. And she was really disturbed. So she came to the call, asking, what's gone wrong here? How can we fix it? And I had a couple of suggestions. The first and most obvious was, one was which, gosh, this might be its whole own episode. In the rendering, all the built ins had been done with a grain. Matched pattern, but with a rift sawn grain, which creates a lot of very densely patterned, vertical striated lines because of the way that it's cutting across the wood grain, this is considered to be very high end right now. It's actually more expensive, but it's not at all a mid-century veneer style. 

35:16
A mid-century veneer cabinet is always going to be maybe quarter sawn, but probably more likely rotary sawn, as in, the long log is stuck up on two pins and a blade is sort of cutting around the outside in an infinitely inward turning spiral. And the result is a wood grain that has more of a topographical map pattern, which shows you, sort of branches and tree rings spreading out more gracefully. It has a very landscapey riverine quality. It looks beautiful. And it's absolutely an object, an option you can request from your cabinet shop today, and she could probably, although I don't know, maybe not, have gotten the AI, sorry, the CAD program that made these interior views from a model to pop that on there, but that would have fixed the problem instantly. 

36:07
The other thing she wasn't liking about them was just how computer generated the images were. They are by nature, harsh, regular, linear. They seem aggressive. They seem decisive. They are giving you an answer, not an option early in the years of my getting this business off the ground, I had a lovely employee who came in. She was a really good designer. Had a great eye for mid-century buildings, but her background was more in a conventional architecture firm environment, and she was used to doing not just all of her image generation, but her design thinking in the program Revit. She wasn't comfortable with sketching, and she didn't want to be responsible for creating hand sketched images for our clients. She thought she could just generate perspective views from inside her Revit models, add a little bit of Photoshop to them, and that would be just as good and much more efficient. 

36:58
The thing is, going back, I don't believe that we can be as creative as designers. I don't believe that I personally, or any designer, can be creative, can be truly flexible, can consider all the possibilities using a program that fundamentally isn't designed for that. It's designed to generate outcomes. It's designed to give accurate construction information. It's not a design program. It's not a creativity program. I wanted to start with sketch, and I have found, consistently throughout my career in residential design, that clients, that people, that humans, respond better to a hand sketched image, as open to possibilities, as giving them options, as encouraging them to ask questions back to the designer, and rather than seeing it as a final outcome, being more willing to mix and match and pull elements from the sketch into others to create the great multiple schemes turning into one a la carte, final solution that is fundamental to the mid mod Midwest Master Plan process. 

37:58
So I was forced to override her and say that, no, even though she preferred Revit, we were going to need to use a consistent model between all of us, and we were going to need to keep sketching as key to both the start and outcome of our design process. 

38:13
Okay, now that is talking a little bit more about the nature of computer as an answer, rather than AI as a problematic answer in remodeling. But I want to talk about coming back to this question of why AI is not a good tool for mid-century research specifically, and that's just because the nature of what we are talking about here, the nature of mid mod remodel, is so specific we're not trying even to create perfect time capsules. 

38:43
It's conceivable, although I don't believe it, that a computer could search for and give you that an AI could give you an answer based on what would be a perfect time capsule, vintage approach to a remodel. But to be honest, a lot of the information about how those things happen isn't available in an online, searchable manner. A lot of the knowledge of the construction techniques and typologies of mid-century homes isn't easily recorded anywhere on the internet. And I know because I have looked for it and so have all of you. 

39:21
It's just not there in blog posts and Revit threads and available for easy AI intellectual property theft, search, scraping. It's in the houses themselves. It's in the knowledge of old contractors and people who do remodeling on mid-century homes and their lived experience. Jumping back to the start of this season, the amount of human ingenuity and discovery and then passed along learned knowledge that is included in someone like Scott Sidler and his life experience of being obsessed with historical windows and then historical homes and just working on hundreds of individual examples and seeing the things that were consistent and the things that are unique between them.

40:05
I just don't it's just not possible for that level of information to be encoded in an AI search. And it really specifically has to do with mid-century and with remodel. So mid mod, remodel, there's two fundamental flaws. So like I was saying earlier in the episode, the term mid-century is so umbrella. It's so broad. We are really talking about something much more specific than that. 

40:29
When I have a client ask me a question about their house, I want to know a number of things before I can answer it. I want to know where it was built. I want to know the specific year it was built, which is going to tell me information that I will process and parse based on, again, human ingenuity and knowledge base, what construction technique was most likely to have been affected. Then we're going to find out. We're going to do a little bit of a hands on research. We're going to see what's going on in the house. 

40:52
And then we're going to process information, not necessarily for what would be a time capsule, but how much we choose to update. What will feel right there is an artist's eye and an intellectual, technical knowledge base, both included in basically every piece of information I give and in the rightness of how it feels to you when you receive it and decide to take that advice or not. Then there's the complexity of the remodeling side. 

41:18
Mid mod, hard to do. And also, as we talked about, the search term mid-century is totally cluttered up by a whole bunch of marketing junk from the last 10 years. And then the remodeling nature is like, this is why I don't think Revit is a good idea to model existing buildings at all. There's too many question marks hidden inside the walls. There are too many uncertainties, things we just can't know or that can't be generalized into a computer world, into an internet world. There are so many things that you really just want to go to someone who is a subject matter expert for. 

41:51
So, okay, fine. What should you do instead? You're not going to ask ChatGPT. Who on earth can you ask? You need to ask someone. Well, absolutely. Ask someone, you know, I will say, generally looking for Internet advice, I have been feeling very much for the last couple of years that I would rather go to a 10 year old Reddit thread for any amount of information than I would take the general advice of AI amalgamated search. I just feel like in this world of ad based, marketing based, corporate based, internet slop, it's so much more reliable to take the word of a person I've never met and never will meet who wrote something on Reddit 10 or 15 years ago. I love that date stamp, and I know that person may or may not be completely correct but is certainly just telling me their unbiased opinion. 

42:46
But I think in the case of making right choices for your house, you do want to get a little bit more than just someone's unbiased opinion. So this is a great place to go to, subject matter specific places. This is a place where you might ask a friend, although, as I have warned many times, you don't want to ask a friend who doesn't like doesn't like mid-century things, but you can certainly ask a friend who cares about mid-century style with you. You could go ask someone who's not yet your friend who cares about mid-century things. 

43:11
And I've talked in the past about how you might find those people online. Instagram account holders of other mid-century homes will happily share with you all the research they've done about their house, all the experience that successes, the failures that they've had in their home remodeling projects, the products they use and what they like and why they didn't like the smell of it. All of that information is as available to you as community minded people on the internet who just want to answer questions. And you know, be have that people please, your tendency that so many of us have, you can probably get a similar level of information with a little bit more back and forth chat inside a Facebook group. In fact, mid mod remodel has a Facebook group. 

43:49
I don't spend a lot of time in there these days, but it's got a very active community of other mid mod remodelers who are sharing their house projects, asking questions, giving answers, opinions, research they've done. These people all care more and probably know more about your house and mid-century houses than the general chat internet. You can reach out in person if you're looking for advice about construction techniques going down to the help desk to the built in customization desk at your local hardware store, not your home depot, but your local, actual Small Business hardware store is a great idea. Your local lumber yard is a great idea for mid-century vintage information. You can check out People to People conversations to be had at your local antique mall. 

44:34
The point is that our best answers are always going to come from inside the community. They're going to come from design expertise. They're going to come from a passionate interest in mid-century design history and how to work with it. They're going to come from human ingenuity, and they are just not going to come from Ai searches and chat GPT. Ultimately, when we think about it, a lot of this ask chat. GPT, rather than what we used to do, which was not a perfect system, but we used to just go Google it. 

45:06
We used to use our own sense of right and wrong our mm, does that feel through I'm going to cross check that. I'm going to consider the source as a way to sift and winnow and judge the rightness of things. At the end of the day, I think it's possible to get a correct answer out of a chat search or AI, but it basically has a stopped clock level of accuracy. You know the proverb a stopped clock is right two times a day. You might get an answer that's correct, but you'll get it in a host of other answers that are incorrect. 

45:40
And you will be not just outsourcing that system and window quality of your own judgment that you get in a Google search, but the sources of the information that it's pulling from will be obscured for you. So you won't be able to judge whether the person who's saying this is a reputable source or not. You won't be able to go look for more information for them in useful ways or not. And at the end of the day, the right choices for your home are not going to come from outside sources. At the end of the day only, you are going to be able to judge what is the right answer, what feels right, what you like, what you can afford, what seems correct. 

46:17
Now, certainly you want to take in outside data to make this happen, but even if you ask me what you should do for your mid-century home, I'm going to give you possibilities, give you options, give you pros and cons, and I'm still going to ask you to answer this question for yourself. By the way, I've gone all the way through this episode without mentioning that I think one of your most handy and reputable sources for information on how to make right choices for your mid-century home is me and other design professionals like me, but we're going to talk about me for a minute here. I have so many different ways for you to get advice about your mid-century home from me. 

46:54
Many of them are absolutely free and just hanging out forever on the internet, or for as long as the internet exists, all of the episodes of the mid modern model podcast, every blog post I have ever made, a host of informative reels on Instagram, some of which are more humorous, but many of which have a lot of data encoded in them, YouTube videos and more, basically, I have just been talking almost ceaselessly to the internet for the last seven years about how to make great choices for your mid-century home, you can go ahead and schedule a design consultation for a 30 minute slot anytime of your choosing on my calendar. 

47:27
You can sign up for ready to remodel and have a regularly scheduled date with me the first Monday of every month to ask whatever questions have come up for you in the last month about what's right for your house, plus listen to other people's questions and answers, or you can work with mid model Midwest to create a mid-century master plan for your whole remodel, answering questions from what kind of materials to choose to how to readjust your layout to what phase of the remodel should happen this year, next year and five years from now. 

47:56
I never get tired of answering people's mid-century design questions, and I just would love to be able to answer yours and know that you're getting accurate, appropriate, actionable data that ultimately you will decide is right for you from someone you can trust. But if you want to start by asking the internet, I would highly recommend that you stick to a good old fashioned Google search or whatever search engine is the most morally appropriate for you these days and stay away from asking AI or chat GPT what to do with your house. 

48:28
Honestly, you might as well just ask your cat. Okay, this has been della Hansmann of mid mod Midwest  hates AI and does not want you to use it to plan remodels for your mid-century home. If you've got opinions on this, I'd be curious to hear them. Reach out to us, shoot me an email, reach out on Instagram and let me know. 

48:48
Have you had success in asking an AI and gotten advice that you found helpful or motivational or trustworthy, or would you also like to rant? Would you like to agree with me that it's a terrible idea. What have you found as another source of useful information in your remodeling journey? Come on and chat with me about it. 

49:09
And certainly we'll be talking about this at our next architect office hours. Call for ready to remodel if you want to join us for that. By the way, it is going to be held on the first Monday of the month, which will be Tuesday February 2, 6pm Central. I'd love to see you there. You can find out about mid modern model, about mid Ma, Midwest, and read the transcript of this anti AI screed, which will turn up in the next scrape for the next large language model, although it won't be found in an instant search, because that's not how AI works, at mid ma midwest.com/2304.

49:46
Stay tuned for next week, when I'm going to be sharing a couple of recent project examples as case studies, but also talking about them as two different approaches to dealing with a time capsule house. I've got two contrasting. Recent Projects, one where we did an amazing, an unprecedented, I would say, job is about to be embarked upon to preserve a lot of the original built ins and quality of a mid-century house, while still making quite a lot of adjustments to it to perfectly tailor it to the new owners. 

50:16
And in contrast, a house that was pretty mid-century traditional to begin with and is getting an entirely new lease on life in a mid-century modern remodel style. So two different diverging paths, you could take two different responses to a situation, and both of them are really great outcomes I'd love to share with you, both visually in the blog post and philosophically here on the podcast. So that'll be next week. See you then.