In this episode of Your AI Injection, host Deep Dhillon delves into the dynamic convergence of generative AI, art, and interior design with guest Gavriel Wolf, CEO of the innovative art consultancy firm IndieWalls. This episode explores Gavi's journey from analog photography to championing independent artists through the challenges of the exclusive art world. The two examine the complex nature of art preferences, the potential of AI in art curation, and the demand for customized corporate art solutions. They then discuss leveraging AI tools like GPT-4 to transform art consultations, streamline creative processes, and shape the future of the art industry.
Learn more about Gavi Wolf, Founder and CEO of Indiewalls.
Learn more about the intersection of AI and art here:
[ Automated Transcript]
Deep: Hi there. I'm Deep Dhillon. Welcome to your AI Injection, the podcast where we discuss state of the art techniques and artificial intelligence with a focus on how these capabilities are used to transform organizations, making them more efficient, impactful, and successful.
Hey everybody, thanks so much. This is Deep Dhillon, your host, and today We'll be talking about leveraging generative AI for art and interior design. Our guest today is Gavriel Wolf. Gavriel is the CEO and founder of IndieWalls, a full service art consultancy that assists clients in discovering and commissioning impactful art collections while managing design needs, sourcing, printing, and framing.
Thanks so much, Gavi, for being here. Super excited to talk art. Maybe we can get started. Tell us a little bit. About your personal journey from, it sounds like you were, had some early interest in photography to become a founder and CEO at Indie Walls. How did you kind of transition that passion for photography, you know, into entrepreneurship?
Gavi: Thanks for asking and thanks for hosting me deep. Appreciate it. So, yeah, when I was younger, uh, probably as in my teens, I got into photography and got into, um. Kind of old school photography and dark rooms and right as the transition was happening to digital photography And I was like digital photography.
It's not gonna be that good like digital photography. The quality is not that good Analog is way better So I still got into darkroom right as it was transitioning and I really liked the the sort of magical piece to it You know the magical piece of the physical creation Um, you're in the room.
Deep: The chemicals, the cool lighting. Chemicals,
Gavi: yeah, the cool lighting. And then
Deep: watching the image, like, emerge on the page, yeah, that's
Gavi: exciting. Totally, totally, and I also really like the sciences and I did a bunch of chemistry in college. I just like, I liked all the pieces coming together. I like the creative aspect to it, you know, like, you took a shot, how do you get it right?
The idea of setting a camera in such a way where you could get an animal moving in the middle of the night. You know, just how you play with movement and light and timing in the physical world. So I really love that. I didn't even, funnily enough, I didn't think of it as an art form per se when I was doing it.
It was just something that I found enjoyable and, and creative and different. Then I went off to college. I got into environmental sciences, chemistry, business, a couple of different majors. And while I was on campus though, I had always kind of wanted to create something. There was some entrepreneurship in my family and I wanted to create a business of some sort.
I had a couple of different ideas, but I was in a. cafe on campus, and there was this artist who had created work. Her name was Alice Mulbeck, I believe, and just had a really cool artwork. It was very thought provoking, um, and I really wanted to buy the piece, and there wasn't a good way to buy it. You know, I could write her a check or I could get out cash, but I was young.
So I didn't have a checkbook and couldn't take out that much cash out of my ATM. Like I have a credit card, but that wasn't the way that it was being bought. And I thought this
Deep: is pretty square and all that stuff where artists just totally lock in their pocket or whatever.
Gavi: Yeah. Yeah, exactly. So there was a cafe called the phone number, go to her, hand her some cash or give her a check.
That was kind of how it, that was kind of how she thought about it. And I thought, you know, it'd be really cool if this artwork in these cafes these creatives are creating could be brought online and everybody could kind of just purchase it online. So that was kind of the original idea and I just liked creativity.
I just was inspired by it generally. I knew I had moments of inspiration. And again, it wasn't like, oh, I'm an art aficionado or I just love artwork in and of itself. I really loved creativity and People that make things in the world. And that kind of just intrigued me getting artwork up in more spaces around the world.
And I'm describing a business that is not the business we run today. Like it evolved since then, but it was the goal of it was a two sided marketplace, bringing artists together with. Public spaces like restaurants and cafes for sale to the public and that was kind of how it evolved. There's always, you know, even while I was doing that, I was working in a lab on producing alternative fuels.
I just liked the physical and digital, right? Like how you, how you put those two together always.
Deep: There's also just always been this challenge for artists to get an audience before they've got, you know, gallery representation, right? They might have their work and sometimes it might have really high quality work, but the art world has its set of standards.
Like, Oh, you know, you're an amazing artist, but you know, you're only 22. Come back to us when you're 52 and then we'll talk. And there's like a lot of that kind of stuff that goes on. And on the buyer's side, there's just like so much uncertainty, like, well, who am I supposed to buy? Am I supposed to buy art that appreciates or?
Is it okay to buy art that's just like on my walls or I just stick up posters of Pink Floyd or like what do I do, you know, there's like a lot of, so I'm wondering if like if any of that kind of art world status quo played a role in how you were thinking
Gavi: about things. You know, I kind of walked into it in a funny way, I think I learned a lot about that as I went along, I didn't even realize when I did this that there was this intense accessibility problem, right?
Not just like, oh, I can't find this thing, like, you know, you need a certain type of screw and you can't find it because it only is at this certain type of hardware store, not even that type of problem, like, there's an accessibility problem of there's, there's gatekeepers, and I didn't understand that when I started, I thought about it more of just, There's cool artwork, and people would like it if they could find it easily, so let's make it more accessible, just practically.
Then as we got involved, you know, we understood much better, and, and there's two different, I would say there's, yeah, there's kind of two different art worlds that, that artists play in. I'm sure there's many more than two, but two that I really think about a lot. And there's kind of that world you're describing, which is this like blue chip art gallery world, where they're like, you want whomever famous person, Johnny Depp, to buy your artworks that everybody thinks is valuable, or the Gagosian, or...
a museum to buy it, uh, or to collect it. And then there's this other world of art where people are creating and they're they're selling it in all different avenues Um, and that's the world that we really exist in right and and that was the world that I got into early It's really independent artists and independent in the sense of they're not represented, right?
They don't have a record label so to speak or a gallery And that was really the world we got into and the world that I really enjoy it's much more In line with with my personality. It's about the work that you're creating and does it resonate with people enough and is it unique enough and different enough and special enough that people are like, I want that.
That's different. That's cool. Just because of that, not because again, some famous person bought it or goes in or some other galleries collecting it.
Deep: Yeah, so, so maybe let's jump ahead to like indie walls. Walk us through, what's your standard kind of customer engagement look like? And what's that life cycle look like?
So a customer comes to you, what are they looking for? How do you like connect them up with the right artist? How do you get that connection? Art over to them. How do you get it? Like the walk us through that whole process. And this is an AI podcast. So along the way, maybe start talking about some of the data exhaust that's happening there.
And that'll, that'll get us all in the back of our minds, like churning away. I'm like, Oh, okay. Like, where's the machine learning and the AI coming in? Like matchmaking with the artists and the, and the buyer. Is it in the content itself somehow? Like, yeah. So
Gavi: a lot of what we do is we are matchmaking that, that is our job.
How do we find. The right artist for the right project. So, uh, right now we're doing a bunch of work for Virgin Cruise Line. Virgin has a very specific brand and aesthetic, and they want artists that jive with their brand and aesthetic. So we're looking for the right artist that fits that aesthetic. And they'll come to us, either a designer who's designing, The cruise, a hotel, an office, we've done a bunch of Amazon's offices.
They'll come and say, you know, we need something. This is our aesthetic. This is what we're looking for. And then we have a team of curators who are really going through. a catalog to figure out what the right artwork is. We're then picking it for them, we're then licensing it or acquiring it for them. We do the printing and framing when there is printing and framing.
Sometimes we're doing big site specific installations, so there's none of that. But if there's printing and framing, we're doing it. And we're kind of handling everything soup to nuts. Everything kind of flows through us. And we make it really easy for the client, and we help the artists get access to these projects they wouldn't get access to otherwise.
So when I had originally started, I actually, I don't know, are you familiar with like 99designs or oDesk or kind of these like matchmaking? Websites, or Upwork, do you know Upwork? Mm hmm, yeah. So, so Upwork's like, you post an open call, contractors pitch to you on this open call, and you can then pick one of them.
So matchmaking has always been our challenge. How do we pair up the artists? And the buyer and do it as efficiently as possible. And I used to try and do it more through a marketplace approach, right? So, or an open call approach, I would say that was, that was part of the idea. We've also used mechanical Turk to try and tag all the artwork properly, such that it was easier to search.
Um, so we've used a lot of different methods. The truth is right now, our main method is just brute force. It's a team of people that really brute force. It's been really, really hard to use. some sort of algorithm or system that's operationalized to pick the right artwork for a given project. Yeah, like
Deep: walk me through what's the, what's the challenge there?
Like, like, is it the, in the articulation of what that aesthetic is? Like, I'm imagining. I don't know. Google seems like a good example. Google's got those big, bright color balls, right? Like yellow, red, whatever. Like, are you trying to, like, somehow capture the articulation of that aesthetic? And then when they, when your search team is going in there and trying to, like, manually find this person, is it because the costs of an error are so high?
and the benefits of getting the right person is important, therefore it's worthy of having humans involved at that level, um, and like, what are they looking for? Like, you know, like, like, what are the kinds of things that humans are looking for to do that match?
Gavi: Yeah, so the prediction engine is really, really hard, and yeah, the cost is high in the sense that You know, I always used to compare it to an NOTA, like, uh, an online travel agent, you know, like, like Kayak or something.
I used to be like, well, everybody would have thought you wanted a travel agent because those were free. And then you can use Kayak, which you, is also free, but you have to do it all on your own. But everybody wanted to use Kayak because it was much better because you could kind of control the process.
But in reality, in the commercial world, people don't want to have that experience. They don't want the experience of searching, right? So if you don't nail it. In a search engine, right? If they put in some, they either put in an open call or they type something into our catalog and it doesn't get pretty quickly the results of what they want.
They're done. They're like, I don't have time for this. That's not what I want to be doing right now. I need exactly what I'm looking for real fast. And I realized that because we tried to push it for two or three years, um, and really tried to hone the system. And yeah, I think just in the B2B world, especially in the, you know, high end commercial world, commercial design world, people want things quickly and they want it to be real accurate.
We weren't able to do that really from a technology standpoint. The interesting thing is I use, um, do you use loom at all? Like the video recording, screen recording, whatever. It's great software. I have a little bit, yeah, but
Deep: I, I'm not a big user of it.
Gavi: So the, one of the coolest AI features in loom that I find is you do a whole video recording and then loom actually titles the video for you and its ability to title the video.
Based on what you said in a five minute video, it will title five words and you're like that is way better title than I would have ever come up with for that video. The AI is really good.
Deep: We actually, uh, have a project running where we're doing a lot of that kind of stuff where we're taking an hour of video, nailing the title, nailing the summaries, writing all the YouTube chapters, all the YouTube summary, all that.
Yeah, it's, it, you can, you can get the machine learning systems, the AI systems do really well at
Gavi: that. Right, so they're really good at that. And that experience is really magical, and it's awesome to hear that you're doing that. Like, I haven't seen it yet there, like, I'll, I'll play with, uh, Mid Journey. And try and just even toy with it to see how close can it get to the vision of what I'm trying to describe.
Like, if you described to me the artwork that you want for your home, I'm just curious, how would you describe what you like in your home? Like, what's your artwork in your home? Or what do you want it to be?
Deep: That's a hard question to answer. We're obsessed with buying art. So we're like, our art is all from our travels.
So we travel a lot, and wherever we are, we're buying from local artists that we meet, talk to, engage with. As far as the aesthetic goes, um, it's easier to talk about what we don't want. We don't care if it matches the couch or not. Like, that's like a major problem. Um, yeah. We don't have giant wall space, because it's just every square millimeter of our house is plastered with art.
So, you know, so size is kind of a thing. But... It's really hard to articulate, you know, like we were in Florence last summer and we were just riding our bikes around town and we kind of stumbled upon this like antique kind of flea markety antique place. And there's like these little shops. And then we just kind of wandered into the shop and I'm like.
This guy's got some cool stuff. So I'm looking through it and there are like old, um, lithographs and I find, and I, and there's like, I don't know, like 500 pieces here. I went in and picked like one piece right off the bat, I'm like, how much for this? And he's like, whatever. It's like a few hundred bucks. I was like, okay, I want it.
And then later on, like, you know, he didn't speak a lot of English. I don't, my Italian is terrible, but we're still talking and he comes by and he's like, well, you know, Macari. And I'm like, Macari is. He's like, Macari is like one of the best, uh, Italian painters. He's like, literally you just bought his piece.
Oh, he's like,
Gavi: he comes back,
Deep: he drops this book off and he's like, you need to know about Macari because you just bought one of his prints and like your eye, you honed in on the best piece I have in this entire place immediately. And I'm like, I don't know. I just, I'm, I'm. I hate a lot of art but there's stuff that I love right away and it just spoke because it's like this it was kind of like trapeze group of these you know these like um acrobatic people abstracted it was like in this really cool pink color and I don't know how to describe it but it was like a good piece but anyway like I have no idea how to describe the art in our house it's all over the map there's abstract pieces there's representational pieces they're super realistic pieces A ton of it's from travels and a ton of it's from friends that are artists that trade with my wife who's like an You know, professional
Gavi: artist too.
So right, right, right, right, right
So so even that whole explanation of the artwork in your house, I wouldn't say it's That, like, I think sometimes designers, or often designers, are a little bit more honed in. But a lot of times it's that, right? It's like, I kind of know what I like. I, you know, I want it to be a certain type of vibe of colorfulness or a certain, you know, mirroring a certain aesthetic of, I want it to feel somewhat industrial.
But then when you get down into it, it's hard to nail it exactly. And I don't know what humans do. And how they can kind of hone and envision, but a curatorial team, they kind of start broad, and it'll be pretty, pretty focused in their broadness, and then you'll start to hear in a conversation with a designer, you're going through the presentation with them, and they're starting to gravitate towards some areas of the presentation.
And that helps the curator kind of hone what that person's talking about. And artwork's just so particular. Like, I don't even, I also don't know what I would like. I was just in, my wife's Colombian, and we were in Colombia. And there was this, I would never have thought that I would have liked, like, a pre Columbian sculpture.
And this guy does these sculptures now, and they're modeled after pre Columbian, sort of, sculptures. And I would have never thought I would have liked it, but there was something about it, there was something about him, and there was something about the artwork, that just spoke to me. And as you see it, you kind of hone in on what you like.
And so I think that, I think AI can do it, at some point. There's something happening in the human brain, as they describe the combination of visual and, and audio that's happening. But I'm not seeing yet even enough where I'm like, oh, I'm going to dedicate a ton of resources yet. But I'm like, okay, I could pull together a discussion with visuals and audio that an AI would be able to pull that together and say, okay, this is the type of artwork this person's really talking about.
Let me, let me pull out more of this, you know? That's the challenge. I mean,
Deep: maybe it's, search requires you to be able and capable of articulating what you're after and that's, that, that can be a problem here. But. What I found is people know what they don't like, and you can give them A, B's, and you're like, Which would you rather have on, like, this wall?
So there's like a lot of sites that have that kind of thing, like, you take a picture of your living room, Yes. The spot, and you're like, A, B stuff, and that kind of interface, it feels like could hone in pretty quickly, and then you could, so you could basically like, A, B, yes, no, no, no, yes, no, no, no, no, yes, yes, yes, and then better, better, better, better, okay, that's good.
And now we can reverse engineer the attributes of what wound up there and the attributes of the space. And now we have like a query if you'll, that can go and, and find it in, in an, in an AI ish context to
Gavi: go further. Yeah. Yeah. I mean, I would say two things, two comments. 'cause the first one was like, you know, it's hard when people don't know exactly what they want.
I actually think kind of, I go back to this loom thing because I wouldn't, and probably the, your software, it's. It's similar in that somebody's saying a whole lot of things, the AI is pretty good at pulling out what the most important ones are of what that person's saying. Yeah. I don't even know how you do it.
You could probably tell me better how it works, but I don't, I don't know. Sure,
Deep: I could talk about it all day.
Gavi: And I'll be curious to hear about it, but the other piece of it, you know, there's a lot of that sort of like ABAB. People lose, it's, even with that, like, I don't like this, I do like this, it's so, artwork is so infinitely variable, and the nuance of it is so hard to nail.
I mean, even, remember, Pandora, like, they were the, the, the music genome, you know? And, and they've kind of gotten it, at this point, Spotify's pretty good. Like, if I pick a song, it's gonna show me other songs. within that genre that I do like pretty well. I still find it like not discovering as much as I would like it like not discovering as well as I would like.
And I think that art is even visual art is even harder when there's multiple medias. I mean it sounds
Deep: like Your business context is one where it's not only about art selection, um, but it's also about, um, articulating the why, you know, like it sounds like you almost have a design team that's like presenting, you know, like, you know, like how you hire a designer to like redo your whatever, and they'll go off and they'll explore this complicated.
Um, you know, set of large permutation of possibilities and then they'll come back and they'll usually give you like One option or two, but never more than three. And the, and the, the, the thing that they're really good at is getting a bunch of like high powered execs to sort of, it's like a toddler, right?
Like, do you want to wear the red pants or the green pants? You got to get them down to a choice space where you're happy. Yes. Designer. I imagine you have a similar problem where you've got like this. I don't know, corporate buyers that are not fully on the same page themselves. And so it's not only about giving them the final option.
It's about prepping that option in a way where, where they've got like a couple of options or maybe a few, but they're happy with it because they feel like you did a bunch to get them there. And if they just sit out there, cause your machine's doing it and then they're back, like looking and flipping and looking and flipping and changing their
Gavi: minds and debating.
Interesting, interesting. That's an interesting thought, that part of, part of them determining that they liked it was actually the process that you went through and the work that you went through, and that's part of how it's getting them to a decision, even though if you had shown it to, I think what you're saying is, if I showed them The end result in the beginning, they might not have even wanted it.
But part of it is they're like, okay, they honed it down for me. We went through this process. Now I feel confident about the decision. Is that what you're saying?
Deep: Yeah. Like, like brand designers, you know, designers that like do logos and branding. They will never. They will never meet with you and then like, like the next time you meet just like boom, here's the design.
Okay. Catch you later Like never that would never ever ever happen because they know Like without all of that story building and then the how did we get there? They're taking you on a journey So you're like totally you're like latching on to them and I think that's part of the challenge with are machines that are so capable of delivering such like poignant insights so fast as we dismiss them because they were so easy to get to it's like when people talk to chat gbd4 and they're like hey you know like deep mental health conversation or something they just ignore it but they don't ignore their own therapist because there's like a social element and there's like a relationship and a rapport and i imagine you're The fact that you're in a scenario where you can afford and it makes sense for you to have like people in that conversation and doing that, it makes sense to me that you would want your buyer in that journey a little bit.
Gavi: Yeah, it's an interesting, interesting thought process. I mean, we think about the journey a lot. We think about it a lot from the perspective of like what we're creating, like we're really about the story that we're creating in the space and we work on telling the story and we do a video with the artists and the designers and kind of creating that whole story that.
how it came to be and what it's all about. But even in the process, yeah, I definitely would agree. I mean, I think there are cases, you know, they're obviously outliers where somebody's just like, yep, got it. Let's move along. Like that's perfect. Let's just do it. Uh, that definitely happens certain types of buyers, but, but definitely a lot of the time it is, it is a whole process.
And, and part of what makes it amazing is that they're like, this was my creative process to get here. Um, and the artwork can be super complex and unique and, you know, it could be like a 40 foot hanging installation sculpture and there's all this nuance in it. So, uh, so there's definitely a lot of that and definitely our goal is that our curators do take people on that journey and there is an experience there.
Deep: usually the case where your buyer has a space already known and now they need a specific singular work for it? Or is it like, you know, we need like 300 pieces to go into this whole building complex that we're building or something like that? Yeah,
Gavi: so, so that piece, like that, that piece I was even describing was for Amazon, but Amazon had 20 different locations and they wanted really local original custom artworks.
Um, and so that, you know, that was that type. I might have a hotel, sometimes a hotel comes to us, we just need artwork for the guest rooms. Same in every single guest room. Two pieces. In each guest room, one, one near the bed, one in the bathroom. Help us find it. Once we do that, then yeah, it's 300, you know, 600 pieces in total, 300 of each piece.
And that's, that's somewhat simpler. I mean, we're probably, like, we can do that pretty well and, and affordably, but that's not where we excel. You know, we excel in the complex and in the story. So yeah, I'm, it, you know, like with so many things, buyers are, I'm sure you have so many different types of use cases of buyers that are looking and what they're trying to do.
I feel like that's often the case.
Deep: It sounds like your scenario though, generally, is corporate buyers. Is that fair to
Gavi: say? Yeah, yeah. Oh, yeah. We do not work with residential buyers. And then
Deep: you're, and your artists are, and they're generally buying existing works or works yet to be created?
Gavi: I would say there's something in between, generally, but, uh, more often than not, it's, um, it's original works, like new custom works.
Sometimes it's like, I really like this print, can I have it cropped in a specific way, or can I have it color changed, or you know, I want it to be horizontal instead of vertical, like there's, there's changes that people will make, or requests being made by the artist, so almost everything that we do is not just like, this piece existed, I'm buying this existing piece exactly as it was, that's, that's less common for us, maybe like 20, 25 percent of the time.
Deep: That's interesting, and like, what's driving that, is it the... The buyers just really want to have, have like unique requirements, like it's a hotel chain with a specific room that's got a specific layout and a specific dimensions or, or is it they want to exert their kind of preferences on the, on the
Gavi: work somehow?
Yeah, and I think this even gets back a little bit to the why we struggle with AI because there's all this iteration happening, you know, it's not just, okay, I found the piece now I'm good. But I think there's two things. One is scale. So most artists don't create scale at the work that That like standardly at the work that we want to create it so they might have a couple pieces but the majority of the pieces are a smaller scale and so This is gonna be a custom piece cuz it's gonna be larger because they're more of the work They're selling is for residential.
So that's one and then two is yeah There's something very particular about their space about the color about the design of it that they want to incorporate in some way We have a client that's working with us and they're like We were working with a different art consultancy, and we saw that they put artwork in a building near ours that's the same, a couple of pieces are the same that's in our building.
They're done. They're dead to us. And, uh, I don't know, totally dead to them, but they were ticked. And they want something that's special and unique to them. They don't want to walk into another place and be like, oh, that's the same thing that I got in my space. Do you have
Deep: challenges with the artists willing to do custom work?
Like my wife always, anytime somebody asks for a custom piece. Her answer is always a retelling of this story. She used to do these, like, huge lithographs in black and white. And they were super detailed, uh, works, and she sold one to this guy. And the guy said, Hey, I love this piece, but can you frame it for me?
She's like, I don't do framing. Yeah, it's like really complicated framing. And then eventually he's like, come on, like, I can't buy the piece. Let's frame it. So she agrees to frame it and she's framing it. And then along the way she cuts her finger and it like bleeds and like a little bit of blood gets on to like part of the, and so she's like, Oh my God.
It turns out she was able to like. totally eliminate it. The guy comes back, she gets the whole thing framed perfectly, and he starts freaking out. He's not looking at the place where the blood got onto the piece. He's looking at a completely other part of the thing, and he's like, this isn't the, the, um, the print that I bought.
And she's like, what are you talking about? And he's like, you know, I mean, I, I, I don't remember this detail. And she's like, look, in 40 years, If you're not looking at this piece and noticing new things, then I failed. So the fact that you're noticing new things is good. So she got so pissed. She just like, she just vowed she's never going to do a custom work.
And so that's, I'm like wondering if you have some challenges there. When you're doing something as subjective as like an art piece, and now you have to like satisfy a group of folks who are buying, like is it straightforward? Or is there some kind of massaging that happens post creation as well?
Gavi: Oh my gosh, I would say a lot of our job is, the reason that companies work with us is to navigate working with artists, and often the reason that artists work with us is because a company like Amazon might say, we don't pay any money up front.
You know, you just got to do it and then we'll pay you after. And we'll say, okay, well, you know, one, we have the capital and two, we have the appreciation of the situation that we'll do it, we'll float it in between when you know,
Deep: Amazon's going to pay, like you, you have a better sense of, I'm not worried.
Artists get screwed constantly. So they're, they're not going to.
Gavi: Right, so, oh yeah, we know how to manage that whole situation, we know how to deal with their finance team, our finance team talks to their finance team, the contracts, the lawyers, all that stuff. So I think that we really navigate the in between, and we try and, you know, we've experienced a lot of different issues on both ends.
We've experienced a situation where the client didn't pay, and they did a bunch of work. We've experienced a situation where we paid all the money to the artist, and then they disappeared. I just dealt with that recently, where an artist was like, Sorry, I can't do this, but I already spent money on the materials, so I'm keeping the deposit you gave me.
You know, so we gotta like, figure out. How to work both of those things and yeah, the, the asking artists to, um, make changes. We're really careful about not committing to a client. Yeah, a lot of them don't like it. Some of them are super happy to do it. They're like, I get to work. I'm like, you're going to pay me.
You're going to pay me 50%. Before I start making any changes and then I'm going to make money on this piece and it's going to go somewhere where people are going to see it. Cool. I'm excited to do it. Something like that. And we're very, we've done this enough that we're very upfront about, I'm contacting you about this piece.
I haven't committed anything to the client yet. We're going to want you to make these changes. We know how meaningful this is to you. If you don't wanna do it, totally Cool. We'll, we'll call somebody else. And I think we just have an appreciation understanding of how meaningful art is. Like this piece that is a part of them that they created is to them.
Deep: You're listening to your AI injection brought to you by xyonix.com. That's X-Y-O-N-I-X.com. Check out our website for more content or if you need help injecting AI into your organization.
So kind of going back to AI a little bit, where, where do you see, if not in the, in the brokering of artists and buyer, where are some areas that you see AI playing a role in
Gavi: your business? Yeah. So two things. First of all, I do see it in brokering. I don't see it there yet. I am very hyper aware of wanting to make sure that I am not the last person to be like, Oh, man, it's one of our competitors or some new company came and they just figured out how to broker it because we started out as technology company.
We're building technology. So I definitely have an appreciation for it. And I think it will get there. I just don't know when because I see how I works. I see how Jackie works and how good they are. It will definitely happen. I'm just haven't figured it out. Like I haven't seen enough of something that would give me confidence there.
So that's one piece. I think the piece A lot of people talk about, well, is art in its traditional form gonna go away? Especially, you know, two dimensional printed artworks like AI can do it. Why would anything happen? Uh, like why, why would artists, why would somebody pay an artist to do it if they could just have it done for free?
I, I do think that AI will be, it will be a new medium that people use in artwork creation. Kind of where the AI ends is where the creativity will begin for artists and. manipulate it will be something new and different and kind of like when printing came out everybody thought that artwork was dead because people were just going to reprint it okay so then people had to figure out what prints were and how those worked and the value of them digital photography changed the way that people thought about photography what did that mean photoshop changed the way that people thought about All imagery that was digital.
Deep: I mean, photography to realism, you know, like realistic painters, everyone was like freaked out. I mean, it's an old, it's an old case and nobody's impressed with something that just pops out of the box and spit something out and can do it repeatedly on anything. Like at some point, art's usually bigger than that.
And kind of more, um, more around questioning and stuff. So where in your business. Do you see some inefficiencies where some automation could really help? So, for example, I'm just saying this as an example, maybe this is not a problem, but like, when your team is assembling candidates for art, they might still be involved in that process, but maybe it takes a couple of days for them to get the right thing.
And then Better tooling to get them kind of more efficiently to the point of, uh, their final selection that they're going to go to the buyer with might be a case where you have inefficiencies, maybe it's in a totally different arena. Like maybe it's in the contract reviewing phase. That would be the place that I, you know, cause like we do this a lot where we just help folks try to figure out how AI would.
You know, it could benefit them, but a lot of times, like one tactic that's really works well is figuring out where you have human bottlenecks, a lot of time energy is spent trying to do something, trying to understand that process more can usually give you clues.
Gavi: Yeah. So one of the things we've been using a little bit.
I don't know if you've heard of gamma, it's a different AI like presentation app. Um, that helps you kind of just automatically put the presentations together. And that's one thing we've been playing with to try and a big part of what we do is put these presentations together, which I always find to be very like silly and annoying that we send PDF presentations to our clients, but that's just always how they want to see them.
Um, as much as I've tried to change the way that they do that, but even that, like how do we put the presentation together so that everything lines up properly right from the beginning, the titles are there, the artwork is kind of laid out in the right way. And that just, like, kind of happens quickly and efficiently.
So, Gamma's one app that I've been playing with. There are others. And trying to do that in a very automated way, that takes, in total of our team, definitely hours every week, many hours of our week. Every week is just putting all the information into these presentations before you're sending them off to the client.
So, automating that is something we've been trying to do. To some extent, one of the things we write a lot about the story of the artist, We're trying to take their bios and reconfiguring them. We're trying to figure out how to write good copy about why this artist is a good fit or what, how they fit in with the story of the property or the story of the town or what have you.
So we have our team using Bard or ChatGPT a bunch to spit out those stories, or reinterpret the artist's bio, or shorten them, so a lot of that sort of storytelling is happening through ChatGPT Bard. Those are some of the ways that we really like to use it. We've been trying to use, uh, like I was saying before, mid journey.
To brainstorm concepts of artwork in advance, like how do we kind of help the client really rapidly get some ideas of the type of thing they want or what the space might look like, etc. That's been less successful. We've been trying with that a little bit. Let's talk about that.
Deep: That sounds kind of interesting.
So it's because you got a lot of control. Are you thinking almost, uh, use mid journey to create an image, like in some imagery that you use as a query to then go find artists that are kind of like that imagery output or something? Yeah,
Gavi: exactly. So, like, you know, we were working on a hotel in Buffalo, and they had some very specific parameters that they wanted, and so we, I don't even exactly remember what the parameters were, this happened a couple weeks ago, but the curators put it in to mid journey to try and narrow down, maybe this is kind of what they're looking for.
It was harder to do, and I think that they could get better at prompting, like our team could get better at prompting, and if they got better at prompting, they might be able to get to a place where they're like, okay, this is kind of the visual that I'm looking for. Yes, this is within the arena. without having to find an artwork exactly like that, that they could then go to the client and say, is this kind of the purview of what you're looking for, and then find other artwork that's similar.
We, we've done like little testing. So we did one where, um, we were talking to a client about if they had wanted a tree installation made out of books, and having the AI visualize what that would look like. And then the AI could say, or the, I'm sorry, the designer could say, that is what I was envisioning.
Or maybe, you know what, actually, now that I see this visual of it, I want the binding on the inside versus facing out. Or I want the books to be different colors or all the same color. I want the books to be cut up or not. Or I want, actually, I wanted an oak tree. Willow, you know? And so I think helping them to visualize what it might look like with the AI, especially when you're doing custom work, is a really quick way to start iterating with them.
We're just in the beginning phases of trying to play with that. I
Deep: mean, that, that feels like a really promising avenue. I mean, we had a filmmaking company that's using a lot of AI on the, on the program earlier, and they were doing very similar things with, um, a set director, like planning out a particular scene.
Um, they would leverage, you know, some generative, generative imagery to like do the sketches and actually sketch out the scene, like how many bar stools are there and like, what does a dingy bar look like? And we wanted a little bit more Western or a little bit more, you know, middle aged or something.
That, that General Avenue seems quite promising. What kinds of problems are you seeing there, getting the imagery out of the generative side so that it feels like it's in the ballpark or is it the marrying it up with the existing catalog of artists and art? I think
Gavi: like several people are probably experiencing right now.
Probably the biggest challenge is getting the people to really. Repeatedly use it until they start figuring out how to use it. Well, people on our team and we're actually doing a, uh, we, we have a company retreat once a year and part of the retreat, we're going to just spend talking about, we're like, we did a questionnaire in advance.
How can we see using these tools specifically in that way? How can people get better at it? And it's a lot of just pushing them to use it until they get better at it. Cause it just takes time. You know, it takes time to use it. One solution to that is. You hire somebody who that's just their prerogative, like use AI to generate this imagery to get this project started and on as many projects as you can.
So I haven't jumped into that yet, but I think it's just, it's not, it's not as it's, you don't nail it right away. I don't know if that's been your experience with, especially with mid journey, it doesn't make like. Yeah,
Deep: no. And not only mid journey, just your whole problem space is a little bit bigger, right?
Like it's definitely requires an iterative process. I think part of part of it is like. It sounds like a lot of the times you have a space, um, you have at least conversation on the buyer's side of the kinds of stuff they want. So, I mean, I might back up a little bit and start with, like, I would assemble a, a dossier on the space and the, and the, and their requirements, right?
So the dossier might require, might have like images and shots of the space, and then there's going to be text as well. So there's going to be people describing. Things like maybe they talk a bunch about something that sounds like a chandelier ish sculpture y thing. Maybe they talk a bunch about stuff, like, on the walls.
But I would try to capture that conversation, capture the general ideas that they're after, and try to capture the, the context and image space. Then I might, like, characterize The visuals in some way, and then from all of that, I would probably start by having someone who's just amazing on your curation team at going straight to solution and then have, and like record exactly what they're doing, like, and get into their mind a lot more to figure out, like, what is it about that?
And I'm going to call that the query. The query is the like space description and the conversation around it. That's the query. And then the question is like finding candidate assets, um, and reducing those down and figuring out like what's going through their head. Typically like designers have heuristics that they're following and they might not think that way or articulate it that way, but that's just the way the human mind works.
If it's like a big old empty warehouse and there's You know, some huge wood beams. There's going to be a rule that fires inside of, you know, anyone who's an interior designer is like, Ooh, I like those beams. We're going to like expose those things. There might be like one person who flunked out of design school who doesn't think that way, but like everyone else is You know, like they might be like, Oh, let's put some sheetrock over that.
But like, that's the abnormal, that's the personnel. And then there's going to be like lighting considerations. So there's, there's going to be rules around that. Like, Hey, do we want a darker space? We want a lighter space. We want it to be warm. We want it to be colder and modern. And like, that's going to drive stuff.
So like that decision. Totally. I think is part of what you want to capture through observation of your humans. And then I'm thinking that if you can articulate it a little bit, then you can start to formulate the transformation into like a more formal query, you know, against, against your imagery kind of stock set.
So for example, like what I would do, like the way I would approach this is probably From two angles. One is that angle, but the other angle is from the, from the artists and asset angle. So like, I would want to characterize the assets too. And I would leverage a bunch of like AI systems on that side too.
So I would want to know, okay, there's like a lot of questions, like how, well, like what dimensions can this artist even produce? Um, you know, like maybe. Their studio just doesn't let them go above four feet square. Cause they live in a little apartment in Manhattan or whatever. Um, so like I would want all that practical stuff, like figure out like what's the volume they can produce.
Oh, it's one person. It's not a team. Um, they do super. You know, uh, photorealistic paintings that, you know, take on average three months to create one. I'd want to capture all that kind of stuff. I'd want to capture their color palette that they're typically working in. Then I'd want to, like, dig into the art itself.
And that's where you can use a lot of generative technology quite well to, like, characterize it. So imagine, like, a really great art curator looking at a piece and describing it. Let's say writing half a paragraph about it, you know, like, what it actually is. Getting the machine to do that. And now you've got yourself a catalog of media assets and artists that you can constrain on.
And all of that is sort of a combination of maybe human well prescribed descriptions of the assets and machine generated descriptions of the assets. And then you've got that high level query that you've formulated from the dossier. And now you have to like, marry the two. And in the early stages, use all your humans as possible to like, Figure out how to art to like describe the work and figure out how to describe the space and desire and then you know You you matchmaking get that sort of process more efficient.
That would be my general approach
Gavi: That sounded really simple. Well, we do this all day
Deep: Literally building a system like this that's not for art for a totally different occasion. Yeah. Yeah. Yeah Perhaps you're not sure whether AI can really transform your business Maybe you don't know what it means to inject AI into your business.
Maybe you need some help actually building models. Check us out at xyonix. com. That's X Y O N I X dot com. Maybe we can help.
Gavi: Yeah, I guess I'm curious, when you're doing this, where do you find it to, like, it works? The most seamlessly and the most effectively. And where do you find it to be that you have the biggest challenges when it comes to implementing AI to like use, create the query and get the data set and put the output.
Where is it most successful? Where is it hardest?
Deep: Yeah, that's, I mean, that's a good question. I think in, in this particular problem, the hard part is that finickiness slash particularness of, of art. Cause it's not. It's not like, give me a picture of a dog on a beach with the beach ball, you know, like that's straightforward.
It's direct it's whatever. So this is more creative, like going back to the Google stuff scenario. Like, I want, um, like, you have to articulate the constraints somehow, so it'd be like, bright, and fun, and colorful. Um, like, there's, there's a million artists that, that Google would never pick, you know. Yes,
Gavi: oh, totally, exactly.
Deep: they're never gonna pick Francis Bacon, like, to represent. You know, the work in their offices, like we don't need slabs of meat and like, you know, bloody faces and stuff everywhere. Like that's not going to go with the office space. Right. So, so like, so, so part of it is like articulating the, you know, like the essence of what they're kind of like about.
And I do think people can like free form that in language. And you could get a back and forth dialogue to like hone that. And, and the reason I think that is because designers do that all day and night long, and I, I bet your, your folks are doing that and they're, and so that, so you asked like, what's the hard part?
I think capturing the intent in a way that. is like queryable is hard in this case. And, and characterizing the media assets in a way that lets you do that is maybe not hard, but maybe you miss some stuff. Like it's, you'd need some iterations there. So like maybe, you know, it's not just about dimensions.
It's not just about, maybe there's like a lot more emotional contribution that needs to be there when, when, when you're training
Gavi: the models, even as you speak, I'm like, well. You know, it would be kind of cool if, if we were able to create a set of questions for a client that was on a, on a phone call, you know, it was on audio that was like, if they answer these questions while we're talking to them and freeform it, and somebody is very good at asking them, continuing to ask them deeper and deeper questions about the artwork that an AI could listen to that call somehow, uh, distill it.
Into what really it is that they're talking about, because we knew the right questions to ask them, because even when I asked you about what type of artwork you like, I didn't get it, really. But maybe if I had a really good set of questions for you that I took you through, maybe as you described it, an AI would be able to interpret it.
really well based on the types of questions that I'm asking you, and it could do it through just the audio listening to it. And I wonder if that could then be distilled enough where it could understand enough to propose the right type of artwork, because, because the query is the whole thing. You have to describe it, that the query is good enough that something That's, that's accurate as to what you like.
Um, yeah, I mean,
Deep: here's, here's like a, this is kind of a fun exercise cause we're just sort of brainstorming and talking about it, but I might start with chat GPT four, like skip Bart. It's nowhere near as good and skip GPT 3. 5. It's nowhere near as good. GPT four is, is amazing and off the charts. And I might start with.
Posing the scenario. So I'd be like, all right, you are an art curator hired by a company to Select art for a particular space first your first task and this would be prompt one Like the scene description prompt is to ask a series of questions that lets you. Oh, yeah Yeah. Characterize the space well. And I might start by saying, ask only one question at a time.
And boom, there's your dialogue. Now you go back and forth. And even with Chet GPT 4, you also say, oh, and along the way, every time you get a response to your question, I want you to update your characterization of the scene. This is a template of a prompt that we use all the time with GPT 4. And it will, and then just try it.
Try it. Take a space. Talk to it, and um, answer its questions, and I'm pretty sure this would work well, like in the, in the room I'm in, by the end of its five or six or ten questions, you would be like, yeah, that, that, that's my space, that describes my space. But that would be the kind of one place I'd start.
And then I'd take that whole methodology and I would apply it to what they want. So you have a finicky art buyer, there's a set of them, they need to answer a series of questions in order for you to characterize what they're looking for. And I want you to first, um, and I want you to generate, and you'll have to kind of go back and forth with it.
And so then, then you'd start. And it would, you know, in like, Maybe you try it out on the, on the Google scenario and it would say like, is there any particular color theme, um, that you like? And they're like, yeah, we like bright colors, bright blues and yellows and reds and, and yeah, and that's it. And nothing depressingly.
Okay. And then it would say, you know, well, what kind of space, you know, like what, what, what kind of emotions, you know, do you find really important, you know, to capture? Oh, well, you know, we want people to be inspired. We want people to be happy. You know, we don't want anything depressing or anything sad or anything like dark, it's through that back and forth.
You can formulate a prompt that will get it to like, understand the context. And I think you should get yourself to a paragraph on the, on the, what they want, a paragraph on the space. And I'm pretty sure at that point you could go back to GPT 4 and you can say, Hey, so assume that you have, you know, some, uh, a catalog of media assets, like, you know, art, um, that's, that's labeled by a set of attributes.
Tell me what set of attributes I should label it for. Like, if you're making a really exhausting search for this scenario, it'll tell you, it's not going to get it a hundred percent, but it'll get it 90%. It'll get it better than any human you talk to. Then you're like, okay, now you can tell it, okay, so hey, um, you have a catalog of media assets that have these attributes that you got through back and forth with GPT 4, and you have this characterization of the space, and you have this characterization of what a client wants.
I want you to formulate A set of keyword queries to query the system with attribute constraints, where appropriate to find the optimal artwork. And then you just see what it does. And at that point, if all that stuff is pretty on the ball and reasonable, and vets with what your designers are doing anyway, now you can go build the system.
If that's what you, if you, if you know,
Gavi: Yeah, yeah, yeah, yeah. No, I like that idea. I hadn't even thought about, yeah, using GPT 4 to do that in that way. And kind of creating the prompt system. But I think that could be really interesting and a fun thing to try and test out. Because I think that is the challenge, how do you create the right prompting?
Deep: I mean like my meta answer to everyone about just about anything is have an extended conversation with gbd4 before you ever talk to me. Because I'm like super busy, got a lot going on, and It's going to answer, like, it's going to get you to 90 percent of what I would say on the topic. So there's no point in me repeating what GPT, it's all like humanity needs to learn how to use this thing at this point.
And they're not used to it and they don't get it. But once you, once you're really good at it, it's, it's just amazing. Like I use it for everything, specking out, you know, specking out products to writing code, to, to, to like, for like writing proposals, like just everything. It's just, yeah. It's not worth doing anything without it, I guess.
Gavi: Yeah, yeah. I mean, I do use it a lot. Like, I use it for a bunch of different contracts that I do. I use it for brainstorming, commission structures for people. I've definitely used it a bunch. But I think I often, the more you use it, the more comfortable you are using it, the more you understand how to iterate with it.
And it's really a commitment to do it and to not... Not be like, oh, let me just go to the old way that I used to do it. So easy when you're not super familiar, just go the old school way. As humans,
Deep: we're evolving really rapidly. And part of that essential evolution is how to interact with these systems. And like, I myself, if I look at the kinds of stuff I'm using these, these models for now versus five months ago.
I mean, it's been a journey. Like at first it was like this fascination with stuff that now just seems trivial that my brain knows immediately I got this. I don't, there's no reason I would ever go anywhere else for that info. Um, but now it's like, it's this kind of thing that I'm describing for you is like a next level up that I've learned that it's not like question answer.
It's like dialogue construction. And asking this thing to, like, listen to me and extract data from the dialogue. That's the template, and I'm using that template all over the place. And there's probably a bunch more that I haven't even figured out yet, but, like, uh, that other people are figuring out, and...
Oh, totally. Yeah, we're all just kind of evolving, because we're, like, adapting to what the machines do better than us, and what we do better than it, and trying to get somewhere.
Gavi: Yeah, totally. Totally. Uh, yeah, it's good. It's good to be pushed along, and it's good to try and think about alternative ways of going through the prompting and stuff like that.
Deep: guess maybe I'll end with one last question. It's always my, my favorite question. If we jump out five years into the future, like what do you think this, you know, your marketplace or art marketplaces in general look like if all this machine learning and AI kind of get somewhere, like how do you think, you know, people are buying art and all that.
Gavi: Yeah. It's so hard. Five years is also so long in this world. I think it will really be about discovery. I think that more so than creation, I think it will be about discovery. And like what you said, it's really about that process of discovery. And I think that AI, I mean I'm sure there's so much that I'm not thinking of, but I guess in my simplest right now view, it seems to me like it'll be that it can really understand who you are, what you're about, what you're trying to do, without you needing to be incredibly good at prompting, but just from other inputs of other things that you're doing.
You pull from the emails you sent, and the drawings you've done, and the Different pieces of information that it can pull from the apps that it's plugged into and be like this is really where I think This project is going. This is this is what I think you should know about This is what I think you should be connected to like so much of the information.
I think Yeah, I think that's where it's really Really helpful is discovering things that you had not discovered before. This is a little bit of a digression But I like backpacking a bunch. I used ChatGPT to to figure out where I wanted to go backpacking within the Northeast And it really helped me to just uncover things that I would not have found otherwise.
You know, just being like, what types of trails? These are my parameters. What am I looking for? And then it started to hone in and I was like, actually, I want things with more waterfalls on them and peaks that are, you know, over 3000 feet. Okay. What am I really looking for? It really helped me Hone in on what it was and discover things that I wouldn't have discovered otherwise.
And I think as it gets better and better that it wasn't even, it was cool 'cause I've never experienced that before, but it was nowhere near the level of nuance that I would like it to have. That I know that it will have. And that was even four months ago, or five months ago. Like it's just moving so fast.
I think it will get to a point. Where it just really understands well what you want to discover. It won't give you the whole thing. It'll give you, this is what you should be looking into and understanding and learning more about in terms of this artwork. Um, I think that's, I think that's hopefully where it will go.
Hopefully we can be a part of making that happen.
Deep: Awesome. Cool. That's all for this episode. I'm Deep Dillon, your host, saying check back soon for your next AI injection. In the meantime, if you need help injecting AI into your business, reach out to us at xyonix com. That's X-Y-O-N-I-X-dot-com. Whether it's text, audio, video, or other business data, we help all kinds of organizations like yours automatically find and operationalize transformative insights.