IDology17

This article is a transcription of the #17 episode of IDology - the industrial design podcast by Mindsailors. You can watch the entire episode on YouTube or listen to the audio version on Spotify, Apple Podcasts or Google Podcasts.

In this episode of our podcast, join Voytek Holysz, Mindsailors' COO, as he engages with the company's co-founder and senior designer himself, Rafal Pilat, and computational designer Jakub Zarzyński. They delve into the fascinating world of AI's influence on industrial design, discussing the latest trends, the power of generative tools, and the ethical considerations that come with this new wave of technology. Tune in to find out how AI is reshaping the design landscape and what it means for the future of product creation.

Voytek Holysz: Okay, guys, let’s get to it. The current AI wave is not the first wave, and it’s not the last one, what’s so special about this current AI wave?

Jakub Zarzyński: Well, I would start with accessibility in the first place because I guess it’s the first wave where AI reached so many people, especially thanks to ChatGPT, I guess. It was like the starting point, and engines like DALL-E for generating images, since, I guess, this is the most interesting one for people when they first start talking about AI. I don’t know if there is something you would like to add about it?

Rafał Piłat: Well, I have been following this for quite some time, I guess, more than a year now. And well, until very recently, the results that you could get with generative tools was so-so and it was not— It was certainly not something that blew our heads off. And now, it has changed. I guess, let’s say, in the first quarter of this year, we got new versions of those tools and the results were astoundingly good. And I guess that’s in terms of both the new ChatGPT starting with version 3.5, and so on. And with tools like Midjourney, and new Stable Diffusion and new models, suddenly generative technology became useful. And I guess that makes this wave exceptional in some ways - certainly differing from the previous generation tools that we had.

Voytek: What’s, I think, also sort of weird that everyone’s talking about this AI craze and new AI tools, when it’s a very narrow field of AI, the large language models and the generative field that everyone is talking about. AI has been around for a long time, it’s been around in manufacturing and in everything that could be optimized from large datasets but what we have now, like you said, the accessibility. First of all, we communicate with it in a different, more approachable way and the other thing is that it generates effects that are sometimes for the masses even fun to play with. So that’s another level of this accessibility. What are those two main technologies based on, like the large language models, and generative AI, is the same thing?

Rafał: Well, large language models are the models that are trained on text, I guess, and it’s an AI that you communicate with by text prompt. So basically, the models are trained on anything that they read, they can be trained on specific material, or they can be trained on publicly accessible materials. So I guess that gives the trainers the opportunity to train those models on very specific material in terms of usability, for instance. So I guess that gives the trainers an opportunity to train those models on their specific material and it would be tailored for their usability, for instance, for a specific company, or I don’t know, medical training or any other example that you can think of.

Voytek: So you don’t need to be a software developer, you don’t need to code to communicate with the software and the output that you get is also not code. So it’s—?

Rafał: Right, you can get an output, which is a code but what is different from previous tools or traditional, let’s say, tools is that you use your natural language to communicate. And that’s basically what large language models are, as another principle.

Jakub: Right now, this is being solved as software for people, so in terms of this accessibility we were talking about, it’s much easier, but it sparks a lot of controversies because currently it is not that much restricted by law on what you can train those AI models. We should also mention that they can be trained on images and they can be taught how they interpret images by text. This is like the next important thing that we should talk about when it comes to the choice of content that you feed to the AI languages and the problems that arise.

Voytek: When it comes to controversy, I think we should mention that a lot of those tools are open source, but they also feed on everything they can find on the internet, no matter if it’s copyrighted or not, no matter if it’s licensed or not, which is controversial in itself. And then people might feed it even illegal content so it can be used in various ways, and being open source isn’t always an entirely good thing. What are the most controversial issues that you have heard about with AI, the current wave?

Rafał: I am sure that partially it’s related to, well, the censorship that’s for one thing because the commercial models that are available and offered by well-known companies, for instance ChatGPT or Claude or beingAI, are actually very restricted in terms of what you can ask them because there are many topics, which are, let’s say, illegal to ask, you cannot ask, for instance, how to build a bomb or anything which might appose…

Voytek: Violate the terms of service?

Rafał: Yes, or be in some way illegal, or I don’t know, gray area of what you do with this information. So of course there are models, which have no censorship, and obviously, those are open source. I am sure that, well, whoever wants to find the information will do it anyway, whether this is something publicly available or will require a slightly more time to get to. I don’t think it will stop people from getting to the information if they are persistent enough.

Voytek: Do those controversies translate to product design? If the models are trained on different artists, who have their own style, who have their own intellectual property, sort of, or even are copyrighted, they are trained on that intellectual property and then designers can use those models trained on those other styles and content to generate their own or not their own designs. How do you guys see this?

Jakub: I guess, it’s kind of a problem. Those models are trained on other people’s work, which is suddenly unavoidable because you have to start somewhere. And I really understand all the concerns about it, because I am not sure if I would be that comfortable if some AI model would be trained on my work. But we had a chance, for example, to use Midjourney in our work. I guess, when you treat your work as responsibly as we do, we try to stay within what you can do with others' work and what you cannot do. We have been using Midjourney, basically, for creating mood boards, for example. So it is like we are generating ideas of other people’s work. But right now, when it’s not so much specified by law, what you can do and what you can’t do, it’s really up to people to put it to good use and use it within some unspoken or unwritten rules not to steal other people’s work. I guess the ball is right now on our side to use it responsibly.

Voytek: But you said, you wouldn’t be comfortable with some sort of AI tool being trained on your work. But what if you would have your own sort of AI tool that you would train for it on your work, but for your use only? Is that something you can see using?

Jakub: I think, like, for the past few months, what I have been doing, I was trying to study all those large language models, and also Python coding, so we can develop our own languages to do that. I wouldn’t have a problem with training that on my work, because it's a whole other different concept because we are doing a tool to be used internally by us. So like, I work for this company, I sell my style to some people. I mean, I use my style to develop new products. So I could as well use it for training the language models and other people’s work from our company. So there’s no problem with that. But I really understand the concerns of professional artists, for example, from illustrators who are worried about their work being stolen to train those commercial models like Midjourney, because they really have no credit from that. They get no money from that and people can easily replicate their style, and not use it as responsibly as we tried to do it in our work. So, there are no strict connections to whatever the output is, it’s just like an inspiration for us.

Voytek: There have been some lawsuits flying around when it comes to copyright and the sort of content that, for example, openAI uses, right?

Rafał: Right. I guess the biggest one was the ongoing lawsuits ChatGPT versus Reddit. Reddit actually sued OpenAI for using their content, which is publicly available to actually train the ChatGPT model on the content of the boards.

Voytek: That’s the key, because those artists’ works that you have mentioned are also publicly available. But still copyrighted.

Rafał: I mean, you cannot copyright a style. I guess this is something, this discussion has been ongoing for some time now. I would say you can emulate a style and to some degree, people have been doing this since art began because, obviously, people are emulating existing styles. Of course, some of the ingenious artists create their own style but then their art style is copied by others or emulated, call it whatever you want, and you cannot copyright a style. And I know there have been attempts to do that but courts denied this: you can only copyright your work. Well, we will have to wait some time until this situation will be cleared. For what I know, the current state is that artwork generated with any kind of generative tools, well, this is at least in the US, the court stated that such artwork cannot be copyrighted. Which means that if you use a tool like, I don’t know - Midjourney, and you try to sell this, you can sell it, if you, make a print of this or any other item or use the artwork to create, I don’t know, print on your T-shirt, anything like this: actually anyone can steal this artwork from you. Well, “steal”. Technically, he or she won’t be doing this because you won’t be able to copyright this artwork. So I guess, this is the ongoing discussion because on the other hand, people claim that even preparing a prompt or prompting the language model is actually putting your own knowledge and expertise to work, and this is also part of using the tools. So obviously, there are voices which claim quite the opposite, that even using tools like generative tools generates a unique content specific to the input that you give to this machine. You cannot get the same result from the same prompt or image that you are prompting with. So I guess, we will have to wait for some time until the situation clears or is, I don’t know, maybe more transparent to anyone. For now, this is still a gray area in many cases.

Voytek: So like, unless you specifically tell the tool to copy a specific style, I guess it won’t be copying a specific style? So it's hard to say that some artists might be in a way robbed of their work?

Rafał: I guess, it comes to the ethics because even if you do this, I don’t know, even if you try to emulate an artist’s creative picture, and then you, try to sell it as a picture of artwork—

Voytek: For example, you made a T-shirt with it, whatever.

Rafał: Right, and you claim that was the original artists, not generated AI. This is unethical. And this is a clear cut case for me that you cannot do this. But people are doing this. Like I said, people are emulating different art styles all the time. They are being inspired by existing artwork, by the best artists, and they are doing this all the time. And does this actually endanger in any way the original artist? I am not sure. So, I guess, no.

Voytek: Probably, their profit in a way. But does this in any way—?

Rafał: But does it really? I mean, if you cannot afford buying the original Van Gogh...

Voytek: That’s an extreme case. And let’s talk about an illustrator whose print you could afford that costs, for example, $200.

Rafał: Right.

Voytek: And you can pay $50 for it. Or you could pay $200 also because it’s not millions of dollars, but still, you ripped him off some profit.

Rafał: But this is, I guess—

Voytek: That’s why it’s controversial. It’s easy to discuss the worst cases. But yeah, let’s not get into it because it’s not something we are gonna solve here. Do you think in your work in product design, there is a risk of a drop in unique designs, originality?

Jakub: Well, I am not really worried about dropping the originality because that’s something that has already started with Pinterest a few years ago. You can see something that is even named as the Pinterest style or trap.

Voytek: What’s the Pinterest trap?

Jakub: I guess, if you go on Pinterest, and try to search for certain designs of many different objects, whatever you are like, at the time, for example - a microphone. You can see there’s a certain style to that. Of course, the images that you will see differ from each other but there are some things that are so important for this Pinterest style that you won’t find anywhere else. There’s something, I guess it’s all started with Apple-like design.

Voytek: Even before maybe with Braun design.

Jakub: Yes, but I am talking more about the internet era.

Voytek: You mean, the Pinterest algorithms choose images that have a specific sort of style to them?

Jakub: Yes. And people put them on Pinterest willingly, I guess. It’s basically a trend to create something in the Pinterest style.

Rafał: I just wanted to say that maybe it’s not a style, sorry, it’s a trend in product design. I guess that’s what makes it stand out. Well, maybe in 20 or 30 years, the next generations of people will judge our era in terms of the style that products were manufactured. We look back to the 50s and those products were designed in a specific way because the manufacturing technology existed which could be used for making products which looked a particular way. And perhaps our era is, I don’t know, marked by this type of trend or style.

Voytek: Flipped, instead of manufacturing dictating the style, it will be generative AI dictating it.

Rafał: Right. I see this more like a trend rather than a style but, of course, I agree with you to some extent, this could be a trap.

Jakub: Right, what I was trying to say is that we should first ask ourselves if using platforms like Pinterest is also cheating because whenever you get an output from a model like Midjourney - I am using this particular example because I guess it’s the most well-known one - whenever you get an output from Midjourney, it’s not something you can use for the final product. Maybe if you are a designer who doesn’t really know about technologies, it is something that you would present to your client, but that client would probably never come back to you. because it is not something that is ready to be manufactured.

Voytek: That’s an important point because it shines a bit light at the— I don’t know if I can put it like that - danger of lowering the bar of entry into creative industries, or lowering sort of the quality because of scale, how many people can so easily start claiming that they are designers like when DSLRs came in, suddenly, everyone could be a photographer, because everyone could afford a professional camera.

Rafał: But could he really?

Voytek: Of course they couldn’t. But still, it crushed an industry in a way.

Rafał: So it’s a question of, actually, using the tool by a specific person. You can give a hammer to a person and he can hurt himself or he can carve out a sculpture of David with the same hammer. So obviously, this depends on the sculpture, on who is operating the hammer. So I guess, this is the same question. If you treat generative AI as a tool, it’s a question of who is using the tool because if you give it to someone who has no expertise on graphic design, for instance, this person will be limited only by this tool, he or she can only generate as much as the tool can output to him. And that’s it. This will be, let’s say, the ceiling for this person in terms of what he can present it with. And obviously, you can already hear voices of people saying that the internet is flooded with the generated images, which have a very specific style in terms of, I don’t know, photography or faces, the proximity of the similarities is very close. When you look long enough on those pictures, you can start identifying certain models which are being used, which are popular. So this is the limit, I guess, for those tools. And if a person is limited, like I said, by their expertise, or knowledge or their skill for using other tools, e.g., Photoshop, Illustrator or any other tools, they won’t be able to do much with this content. This is going to be the stop for them. And obviously the entry bar will be lower for many people but it is a question of who is actually a target audience for those artworks. If you have a budget of $5, then you can, sure, buy generated content if it will suit your needs. But if you want something tailored specifically for your needs, then you will have to hire someone who is actually an expert in this particular case.

Voytek: How does this translate into product design? And what sort of AI tools do you guys use in your design work?

Jakub: I guess we can obviously start with the ChatGPT because it’s helpful and I don’t want to say alternative to Google, but it’s pretty much helpful whenever you are doing any research. Maybe, it is not so important for the visual style or visual aspect but apart from that, we have been using Midjourney and Stable Diffusion for different projects. We spend a lot of time using both softwares so we pretty much know a lot of their limitations. Also, we have learned, when in our design process we can use it so it benefits us and it’s not just a novelty or a gimmick. And that’s helpful. I pretty much liked when we did, because usually when we do mood boards for our clients, we use all sorts of platforms, of course, for example, Pinterest, but it was pretty fun to put some Midjourney outputs, among all of those inspirations. Rafał, you are a specialist when it comes to Stable Diffusion. So I guess, you could say something more about this, because it’s really interesting what you can achieve with Stable Diffusion.

Rafał: I already talked about it with Anita, but we are using Stable Diffusion in a few cases, when we try to automate making sketches, for instance, that was a very useful tool. We used screenshots, raw screenshots from Solid Works, just having the outlines of our object, and we wanted to generate sketches, which were resembling naturally handmade sketches. We had quite a number to prepare and we wanted the style to be, let’s say, matched across the different sketches. This is where we use Stable Diffusion for and it actually has cut us a lot of time. And we basically didn’t have to do this manually. We generated around 20-30 sketches in about 10 minutes and that was really something. And for other purposes, Stable Diffusion is great at imagining specific artwork. You can control it very well by adding various levels of control nets to the process. And you can, let’s say, process a cartoon image into a realistic or photorealistic output. And we use it to transform our semi-realistic renderings of people or characters into photorealistic persons and this is a great workflow which takes our work even a level or two up in terms of the quality or a realism when it comes to people visible in the pictures.

Jakub: I would like to add one more thing that is a different field: to find suitable workflows between different softwares. We are far past the era where you are specialized in only using one software and that’s it. What we have been working on for the past few months, when the AI really started to shine, is to try to develop different workflows between different softwares. And that was quite an interesting journey for us to find those solutions on how to really use it to our advantage.

Rafał: It was a challenge because when we wanted to use a specific function, we had to turn out that we had to learn something new, in Blender or learn a new software to create characters, for instance, that we used. We benefited as a team because our knowledge has increased in terms of using various different tools. Given that we have AI as a tool as well, this gives us a good head start or a good advantage over our competitors, who don’t use those tools, and I guess, that makes this challenge even funnier.

Voytek: Outputting 20 or 30 sketches in 10 minutes, that’s a high pace. Is it referencing our topic? Is it cutting corners? Is it cheating? Or is it the evolution of art?

Rafał: It’s time-saving. And in the end, it’s also money-saving. When it comes to making those sketches, we would have to decide who will do it in our team, or we would have to outsource those sketches if we didn’t have enough manpower to handle this. And obviously, we can, by automating certain tasks, we can do it and still have control over it. And this saves us time, gives us more time to do other things, or it gives us more free time, because obviously, we can use this time to train on Python or anything else, which we are currently doing. So it’s a time saver.

Jakub: I would also say that not only is it a time saver, but it is really like your third hand sometimes. Because when you are a designer, and you like to use sketching for your work, you sometimes take, for example, a blank, A3 paper page. You do all those sketches, like you don’t start with the photorealistic renders, you just do the basic napkin sketches. So, you do a lot of them in a small period of time anyways. So, we will end up with, for example, an A3 page of your own sketches. But when you are doing them, at the same time you can put some ideas or some of your sketches that you already did into this Stable Diffusion, and while you are doing your sketches, it can do even more. It’s like another worker in your bureau sitting by your side. And just using something similar to your style so it basically becomes a conversation, like you do with ChatGPT. But it’s more on the visual style - it helps to take a new perspective on your own work.

Voytek: That’s an interesting approach.

Rafał: I really love remixing things in either Midjourney or Stable Diffusion. And as a result, within an hour, I can generate about 60-70 different images, output images, which I treat as an inspiration for my actual work. This gives me a better starting point rather than browsing through Pinterest.

Jakub: Where everything looks more or less the same.

Rafał: That’s right. And sometimes I am really surprised by the randomness I get from those images. Sometimes they are totally abstract but sometimes they are pretty useful. And I guess, they could indeed serve as an inspiration to generate an actual design from this artwork. So, I use it as my private inspiration tool for finding a specific shape or exploring— Doing an exploration journey, before I actually seek to design something.

Jakub: I think people are usually worried that the outcome from all those engines will be the end result. So basically, anyone can generate something that is a final product in a few seconds but it’s not, it’s basically just the beginning of the work, whenever you get those outputs. If you are a professional, and you look at this, you know, how many more steps, how far it is from a finished product. When you dive deeper into the whole AI wave, as a designer, I would say, you basically realize that it’s something that can help you, it’s not there to make your work obsolete or take you out from the market. Of course you have to adapt to those things. If you are still sticking to only doing sketches on paper, not using Photoshop or other things, you are basically not up-to-date with the technology and you have to be.

Voytek: I think that’s a great summary for the whole conversation. If anyone is afraid or thinks that using some of those tools might be unethical or unfair, I think what you just said is the crucial point. You need to be up-to-date with technology and you need to use those ways. We need to use those tools in an ethical way, like you said, for your own inspiration, for kickstarting your creative brain. Never for final work, basically, unless maybe you train it on your own design. Closing up, what do you say, in your own words, to anyone who is a product designer and says, “I don’t like, I think it’s cheating.”

Rafał: Don’t close your eyes and pretend that those tools don’t exist, because your competitors will use them to their advantage.

Jakub: So they are already using them.

Rafał: They are already using them. Don’t pretend that this doesn’t exist. The evolution is ongoing and I guess when you stop learning new things, I guess maybe it’s time to change your job and not pretend that this is not to try to enchant everyone that they are cheating.

Jakub: One more thing, I would say, if you are on the side of a barricade that is worried or it sparks a bit of fear in you when you see how good looking the outputs could be, just try to read more about how those engines work. It is like the simple case of “the less you know, the more you fear”. So whenever you start to learn about it, you see how it can benefit you and it’s not really that scary, if you know what you are working with. Don’t be scared to gain additional knowledge about the topic.

Voytek: That’s great. Thank you guys.

 

kuba.webp

Jakub Zarzyński is a versatile computational designer specializing in integrating various design fields to find innovative solutions, with a focus on utilizing new technologies and collaborating intensively with diverse softwares.

 

 

 

 

 

rafal.webp

Rafal Pilat is an entrepreneur and co-founder of Mindsailors, an awarded industrial design company, with over 18 years of experience as a designer himself.

 

 

 

 

 

 

Wojciech-Holysz.webp

Voytek Holysz is the COO of Mindsailors with 16 years of experience in running a business in creative B2B services, marketing, sales and video production.

Let's talk

Schedule an initial talk and get to know us better! You already have a basic brief? Send it over so we can have a more productive first meeting!

Contact us
Set up
a meeting

This site gathers statistical data in order to enhance user experience and to improve the content we deliver. We never store any of your personal data. You can read more in our Privacy Policy.