Alexia: So maybe as an intro, it could be really great. If you could introduce yourself and tell us more about what are your main interests in your studio, what type of project you do and all that.

Rifke: I’m Rifka Sadlier and I’m sort of one half of the XR zone, which is. I suppose it’s a joint creative practice, but we kind of, we ended up calling it all sorts of different things. So sometimes it’s , oh, we’re a digital design agency. Other times. We’re , we’re two creative technologists. Mainly we make websites that we kind of do a little bit of light designing branding, art direction outside of that. Our main focus is kind of on the web. We’ve kind of. Sort of just by accident, falling into making websites for the arts and music and fashion, which is really fun. So that suits us and they’re kind of, I think they tend to be the clients who are happiest to be quite experimental with stuff. So yeah, I’ve kind of, we’ve kind of ended up doing a lot of sort of 3d or slightly experimental websites and occasionally. Yeah, more recently, I suppose, , because machine learning become such a widespread thing and it’s become much bigger and kind of exploded recently. So we’ve kind of been in touch with people who are kind of working in those fields. And I think a lot of our clients kind of want to bring that in because it’s current and new and they want to kind of try it out. We’ve kind of had a few interesting experiments, mostly with people working other people working on that side of things and then kind of being privy to all the conversations around it. And then a few where it’s kind of , we’ve had to use tools to kind of figure a little bit out ourselves and it’s been a real, , Just scratching the surface, but still , quite exciting, , oh, it’s doing something by itself.

Alexia: Thank you for the intro. So before we diving into more machine learning specifically could you maybe tell us, , as a designer, what’s your relationship to tools? , is it, a mean to an end or does, or do they play a more important role in your process in general?

Rifke: Oh, that’s interesting. I think, I suppose it’s a means to an end in a way, but also I know that , I , cause I studied graphic design and kind of came from a design background and then fell into coding. Cause I was , oh, just pick this up as a side hobby and then found that I really enjoy doing that. So now I think it’s , I’m very comfortable with the tools I used to code stuff and it’s a bit more of a wrench to try and move back on to sort of visual design software. So yeah. I guess. Yeah, I guess it is a means to an end, but I think my practice is informed a lot by being , oh, the nice, comfortable text editor that I can go back to. And I know how it works now. So I think, and I think also with coding, because a lot of it isn’t , it’s not kind of just happening, visually there in front of you as you’re working on it. and I mean, it is to an extent, if you’ve got two screens and stuff, but it’s, there is that kind that’s slight removal where you’re writing something that has a visual output. So I think a lot of the time the tools do inform, what happens? Cause it’s you do something and you’re , well, what’s going to happen. And then it, , I don’t know either it would be what you expected, but most times it kind of necessarily be exactly what you think it’s going to be. And then you’re , oh, okay. , we’ll go with this anyway. Or, or that kind of informs another idea or something that, I suppose. Yeah. So, yeah, I think a bit of both.

Alexia: The next question is more related to machine learning, but still in the realm of tools, could you maybe share with us, what type of tools using machine learning did you use before, and maybe tell us more about what was your experience with those tools?

Rifke: So I’ve used a bit of ML five. There’s another thing, actually. I always forget that this is machine learning, so it’s, so it’s so simple. It’s 200 lines of code. That’s just a really simple face detection algorithm. So I’ve used that a bit. It’s called Pico JS. P I C O. Yeah, that’s mindblowing. And then what else? I suppose, Spark AR, I’ve used a little bit and I suppose actually, that is kind of based on machine learning. I suppose a lot of, a lot of things are, but it’s just , so kind of under the hood that you don’t think of it as working with machine learning. And then as I can’t remember, it was one of the Google platforms, that my friend used when we were working together . And it was for the hip hop poetry book project that he trained on a dataset of lyrics that are kind of , they’re still being , contributed to now. And you’d kind of give it a topic and then. Give it a. Prompt. And then it would kind of write your own kind of poem based on this data set of lyrics, which is amazing. So that, I think, yeah, my experience of that was , oh my goodness, this sounds so stressful, but it’s , it’s so interesting. And the output is so exciting, but it’s actually also so difficult to make the, especially text generation models. It’s so difficult to make them behave and not say something awful. And I think it was GPT 2. So it was. I have way less context than GPT three and said so much dreadful stuff. So I think seeing that it was , oh my God, working with machine learning is just a process of trying to stop it from saying bad stuff. Okay. , that’s stressful, but still quite cool. And then I’ve had, , I dunno, I’ve had a little play around with with open AI’s little GPS playgrounds. I’ve not, I’ve not trained it or done anything with it. Yeah. I’ve spent hours just in front of it, typing stuff in and.

Alexia: Yeah. It it’s quite fun to use them. And we, you know, we did a workshop last semester with GPT 3 and our students. And still, it said some pretty horrendous things. All students created their own data set and everything. Cause they’re hard. It’s hard to control still in a way. But I think it’s interesting. This experience with all those different tools. , could you maybe tell me did you feel limited somehow at which stage of your creative process, did you use them, for example?

Rifke: So I think, oh, for the most part, , I suppose for the most part, I was kind of doing the sort of front-end coding and just, and that was kind of going on in parallel, but it was. I suppose the limiting aspect was , oh my goodness, , this sounds stressful, but there’s nothing I can do to help. And also if they can’t find a way to tame the machine learning algorithms, then the project won’t go ahead. But I think, but yeah, it was kind of. Yeah, I think when I’ve tried to use it myself, it’s kind of the only real limitation is time and my concentration span and my ability to actually absorb all the information you need to know to in order to lay, actually use the tools, how you want to use them. So yeah, I still haven’t really, I mean, For ages, I’ve been , yeah, that looks really cool. I want to , be able to learn how to use it. And I still haven’t really done more than scratch the surface. Just cause it feels a little bit , oh my God, it’s this huge thing.

Alexia: It’s a big, big thing to, to dive into, but it’s interesting that you said stressful what’s different. So stressful about it is the fact that you can’t control the outcome numbers or.

Rifke: I think it was that they were , there’ve been a few projects where they were just so many kind of calls and stuff that everyone is part of. And then it ends up being the people doing the work on the machine learning side of things, being , oh my goodness, back to the drawing board again, this is it’s, it’s saying, well, if we’ll stuff again, and , we’ve tried this, this, this, this, and this. I think in the end we had to, at first we were considering using a list of words to block. And in the end rehabs to use another machine learning based tool, the oh, what is it? The it’s an API and it’s called something I can’t remember, but you might know it that it kind of detects toxicity and statements.

It gives you a percentage. So I think there was that. And then also there was , instead of having a list of words that were blocked and we had a list of words that were allowed, so it couldn’t say anything outside of, , if it said a sentence with one of those without, with a word that wasn’t in the list, I’d be , no, you can’t have it. But then also, obviously we’d the context of things and because yeah, just contextually, . You know, slang and stuff moves so fast. So it would still , even with the list of allowed words, it would still come up with some stuff that was , oh no, it said something again.

Alexia: But I do now that you have all those experience and also know the frustration of, you know, working with algorithm, , would you say that you’d be interested in having those tools being part of your creative process?

Rifke: I think definitely. And I think for projects where it’s not kind of so important that it doesn’t say something dreadful, so for projects, it’s for a smaller client, or it’s kind of clear to users that is something really experimental rather than: this is what we stand by and believe in exactly what the thing is saying.

I think, I think the fact that it’s so chaotic can be really fun and kind of . It can kind of be part of the work itself, being , oh, look, if you teach something loads of bad stuff, or if a human trains, a machine, it would be horrible and biased, the human. And I know that’s kind of points been made quite a lot, but I suppose it there’s a reason for that. And it’s, cause it kind of does brew. But yeah, it’s kind of. It’s the chaotic ness of it almost made me kind of more drawn to it, to use experimental and personal projects.

Alexia: You talked a lot about more texts that generated texts. Have you ever tried it with images or? I don’t know what other type of I mean, I’m thinking about sound, for example, I’m thinking even those AI tools that allow you to select sketch to code and they would try to before, but apparently you can sketch it. And it does the HTML structure, which I think is.

Rifke: I’ve tried, I’ve only had a little go on the, I think it was on ML5. They used to have a little, maybe they still have it. I can’t find it anymore, but they had this little playground. It was , you can throw something. Give you an image of your drawing in the style of something else. So you’d , it was draw a cat or draw Pika Chu, and it would come out with this really horrible looking, sort of , not quite hyper-realistic, but sort of uncanny valley, horrible little image. So I’ve never, I’ve never got to the point of actually training something that. I’ve kind of late. I spent hours just being draw a cat. Oh, it looks horrible. Amazing screenshots at drew another one. But yeah, I’d love to kind of, I think also, cause a lot of those, these kinds of tools are kind of they’re there and you can just use them. I’d love to find a project where I can apply it just for yeah. It’s some kind of fun interaction for people.

Alexia: And then it kind of drives me to the next question. This is more in the, in the field of graphic design, but you can also find other examples of course. What type of machine learning application do you think right now are underused , or are there things that you would be really curious to try?

Rifke: I think. Wanted to , have a play around with sort of image making GANs and stuff more. That sounds really fun. And obviously, yeah, runway, I think this wasn’t me, but my business partner, Dan did , he, I think it was runway. He used anyway. He put in , Just a huge dataset of pictures of the insides of pubs. And then got it to kind of come up with its own idea of what the inside of a pub looks like. So that was really fun. I was , ah, I’m jealous. I wish I’d thought of that, but that kind of made me think , oh, I want to have a go using it,

Alexia: but then what’s the right now. It’s just, do you have find the project that makes you use those tools are there any limitation to you that you feel. you are not trying those tools yet because technical stuff, side of it or other things.

Rifke: It’s the thing where you, , when you get into anything new where you’re , oh my God, that’s just so much to learn. And especially in the world of tech, whenever you get into something and I remember this, just getting into coding in general in the first place, , you’re just , why on earth do I start? I think it’s one of those ones where , I’m hoping there’ll be a really specific project where it’ll be , okay, there’s a really clear use for this particular thing. And then kind of work my way around. Because at the moment, I’m kind of , oh, that’s also cool. I want to know all of it, but I can’t learn all of it at once. And then you kind of , just flip between 10 different things and don’t learn any of them properly. So yeah, I’m kind of waiting for the thing where I’m , okay, I’ve got an application for it now. And I’m going to. Just try and find out this very narrow thing first.

Basically, and also that kind of imposter syndrome of , oh no, I couldn’t do it that’s the difficult ones. That’s not for me.

Alexia: Yeah. But I think it’s, I mean, tools runway makes it. I’m sure that if you had to go, I runway, you will master it pretty quickly, but they’re definitely here to make it more accessible, to make machine learning more accessible to creative. So the definitely been, you know, amazing in that and making us less scared, I guess, to play around with the machine learning.

Then I have more question let me, let me have a look bit more related to what type of new I’m rethinking of graphic designers. And it’s interesting because you have a background in graphic design and now you’re into coding and you building websites and all that. But what type of new skills do you think graphic designers need to learn to be able to work? Do they need this to be knowing that we have more and more automation?

Rifke: Oh, I suppose, , I suppose it will become kind of less the sort of jobs that are kind of more repetitive and automated. And , I suppose, , there’ll be less need for art workers, for instance. And kind of more jobs kind of instructing and directing and putting the input into the machines rather than being the person who’s actually executing it. So, yeah. So I wonder what that will do. Cause obviously that’s always , when anything gets automated people , oh my God, it’s all becoming automated, but then sort of, you know, the industry and society does adapt around it and then more people are managers or art directors or, and then there’s just , there’s also , because things have been sped up by automation, there’s also more work that’s created by that. And whether that’s positive or negative, , I don’t know, but it kind of. Is so I guess, yeah, it would become more maybe graphic design will become more conceptual rather than about draftsmanship, which it already has from , you know, Even 10 years ago, I suppose. But yeah, I guess I, I guess it will carry on going in that direction,

Alexia: but there’s something that I want to go back to that you said I thought was interesting is where, where was it where you said machine learning has this chaotic energy, how would it feel to it? And I think that’s interesting cause you you’re the first one. Mentioned it that way. And in a way where, what I find interesting is that suddenly it’s not so much a tool that here to assist you into those automated, you know, really simple task in a way. But more the unpredictability that can come out of it. Could you say more on that?

Rifke: I guess, I don’t know. I kind of almost think of it. , I don’t know while I was , oh, machine learning, scary. Cause it’s kind of almost it brings up questions of , you know, what is intelligence? What makes humans different than machine? , is life just meaningless? Oh my God. I’m , God, I feel existential. This is horrible. But it’s kind of. That’s this whole idea of it being dystopian and people being , oh, it’s the machine, it’s cold. It’s horrible.

But actually more recently I’ve seen loads of works coming up where people , say there’s a book by K Allado Mcdowell they’ve said , oh, the, you know, it’s a collaborator rather than , this is a machine. This is my work. Does this weird shifting relationship with machine learning that is starting to happen where it’s , we don’t see machines as this scary, enemy that’s going to take over. I don’t know the way people talk about their Gans or their algorithms. The models they’ve trained and stuff. It’s people talking about their babies and it’s kind of , it kind of is a bit that. It’s if you raise a child and you’re really horrible to it, it will probably turn into an adult who sees the world as like this horrible place and act accordingly. And it’s kind of the same with machine learning. If you , what you input into it, it is kind of chaotic. And before you kind of fully train it up, it will say some really ropey, out there stuff. kind of like a toddler or something.

But if you, yeah, if you kind of input a rounded sense of the world and what it is like then it will grow into something more well-rounded and reason stuff. So it’s kind of interesting and it’s , does it really matter that it’s so different from human intelligence in that way? Or , is that something to be scared of? We could equally just be scared of people raising children in , horrible, biased way, which people already do.

Alexia: I mean, I, I love that you said babies, , I don’t know if you know the world, the work of a Holly Herndon she’s just this amazing musician, not some professor based in California. She basically created this AI baby. that sings I’ll just drop you a link of some her work, but it’s really interesting that you said baby and this new type of relationship almost that we can build, with with those, those algorithm

. If you were to, if we’re pushing a bit in this idea of collaboration outside of automatic daily tasks, , what would you to do with a machine learning tool?

Rifke: That’s really interesting. I think , oh, that’s really difficult. Actually. I think that’s also one of the reasons I haven’t started using it. I’m not sure what I want to do. I want to do everything. I don’t want to do anything. But maybe working with, I think working with sound would be really interesting with it. Just cause it’s an area that I don’t really know much about. So I feel the output would be more kind of , oh, that’s a surprise. But yeah, I think also, I don’t know. I’ve seen who’s at my mind is so blank today. What’s her name? She does computational, , machine learning, generated poetry.

That’s the one. Yeah. Yeah.

Alexia: It’s beautiful that the work,

Rifke: It’s lovely. And actually she’s someone who really talks about machine learning is a kind of sentience collaborator. And I think. I watched a talk by her fairly recently. And it kind of, and that I think was one of the things that kind of made me be , oh, maybe machine learning, isn’t something to be scared of. It’s something that you can be , oh, the machine it’s , come up with something beautiful. And , it can just be as beautiful as written by a human. And you can give its output as much respect as you would that of a humans output. And it was really interesting. You. I showed my mum, the talk afterwards, and my mom’s kind of she’s in her sixties.

She’s still , oh, the machines I coming to get us, what’s going to happen. You know, she’s , doesn’t holding her phone to her ear. Cause she’s. The waves, the waves, the radiation. So she’s quite sort of technophobic and in a kind of , actually scared of technology rather than bad, it’s it kind of way. And she was watching this talk and I was , yeah, see, it’s, it’s demystified it. It’s really nice. It’s lovely. The machine’s lovely. And she was , no, I don’t. Oh, , and she still couldn’t get her head round it. Cause she was kind of , oh gosh, it’s desecrating poetry, poetry, poetry is what separates us from the machines. And I was kind of , oh, maybe, maybe we need to get rid of the idea of , trying to separate ourselves from the machines. Machines aren’t humans, but they are kind of extensions of our own minds because that’s what, , what comes out with them is a kind of interpretation of what we’ve put into them.

They don’t exist kind of. Independently of us, but. Yeah. Sorry that didn’t answer the question at all. I just went off on a tangent. No, no,

Alexia:

Actually, maybe we shouldn’t think that this separate entity and what if the machine was more continuity of us? Rather we agree on I’d you don’t have, I think that’s definitely a really , interesting idea. Then what do you think of something I wanted to have your opinion on is, you know, we still, those gun images, imagery generated the aesthetic finance really present.

They almost became, I think in a way, not common that we’ll be pushing it too far, but we’re more used to it now, do you think, I mean, do you think there’s a risk of having homogeneity aesthetic by using machine learning and GANs, for example.

Rifke: I dunno. I think it’s , I think it’s one of those things where it’s , it sort of reached that point of , oh, it’s the thing to have some Gans generated imagery on your poster for your night. I think when people have got over that, initial kind of , ah, Gans generated imagery, then it’s going to kind of go to that point where it’s , people will find the actual kind of use for it and then it will become, you know, it won’t just be that all the images in the world is suddenly Gans generated. It will be , get over the sort of honeymoon phase of it. And then it will become more a part of daily life. But I suppose in the way that any new technology is , there’s a big boom in it. And then that’s kind of a bit scary too. You’re , oh my God, is this just life now forever? And then it’s , then it kind of settles down a bit and just takes its place and sort of everyday life, a bit more. And

Alexia: I just have one final question to wrap it up already. What do you think would be needed for you in your mind to to get those tools machine learning tools more accessible to designers? And also maybe young designers know that just starting their carrer

Rifke: I I kind of feel when a big software company, Adobe kind of, I don’t know if they’ve already got something actually. Perhaps they have, but

Alexia: Adobe sensei is kind of, it’s more of a for marketing and optimizing images and, yeah.

Rifke: Okay. I suppose, if, Adobe brings out something that’s Adobe GaN or something that where it’s , okay, this is the kind of. The thing that we’re all told to subscribe to and downloads, then I think having it kind of just come become part of the set of things you’re expected to use. We’ll probably open up a lot, but then also Adobe is kind of. It has got this kind of exclusivity thing, so you have to pay for it. And it’s not fully accessible to people who can’t afford it. So I think just online tools and , especially things with visual editors rather than cause I think, yeah, the thing that scares a lot of people off is it’s , oh, you have to code it and train it. And it’s all a bit kind of in a black box and a bit just mysterious. I think that’s probably off-putting to quite a lot of people, but yeah, I think, I think also things spark AR I’ve kind of opened it up a lot. Although there’s still this really big divide, a lot of people who haven’t used it before, just assume it’s this really difficult thing to use. Whereas it’s actually , a lot of people have been , oh my God, this is actually really easy. And it looks really impressive. I’m going to use it. But then there’s another half of people who are , oh my God, Did you know, that person makes filters and you’re , yeah, you could make a filter really easily. You don’t have to teach the machine yourself. It’s literally fine. And I guess, yeah, I guess more kind of software where it comes in, a package, neat package kind of thing. And it’s clear that you don’t have to configure it or But yeah, I’m guessing , cause a lot of stuff’s already come out, but it’s kind of still , quite , whoa.

Alexia: So then in a way you, would you be saying that if, if you were to use machine learning in your creative process, would that be more at the beginning to generate ideas possibilities or will it be ready to reach a final outcome?

Rifke: That’s really interesting, actually, I think, yeah, I hadn’t really thought of using it for idea generation, so probably to reach a final outcome and kind of be , usually I’d be , oh, it needs to be the interactive bit, because a lot of my, most of my work is kind of interactive and it’s about the end user, interacting with whatever it is. And I think, cause I used to let forwarded graphic design sort of for a few months studied illustration. And then I was , oh no, I’m going to move. But it was , everyone was meant to keep a sketchbook and do loads of prep work. And I was , no, no, no, no. I just want to draw the final thing.

And I think I’m still quite that with work. I’m , I don’t want to do loads of prep at work, which I probably should because that actually leads to better outcomes. But But yeah, I imagine using it as the, the interactive bit that someone comes face to face with, I really the idea of kind of using it idea generation and as a collaborator and stuff that.

Alexia: Yeah. To be, to be continued then. Well, thank you so much.