This transcript is generated with the help of AI and is lightly edited for clarity.
//
JONATHAN:
I had generated 40,000 images in Stable Diffusion at some point. And the reason was, I was trying to make my own RPG. I had designed the book—you know, my background is in product design—but when I laid out the book, I realized I needed 45 pieces of art and I didn’t have it. And so I turned to Stable Diffusion. This was back when it couldn’t do faces or hands that well. I could draw faces and hands, so I would draw over those and make the characters. And it was really fun. It’s just like an idea generator. True art, right? Creativity is improbable.
//
PARTH:
The token grantee program gives $1,000 a week in tokens to high-potential creators already deep in AI, across film, gaming, comics, print, and digital art. There are no tool restrictions. The freedom to choose is the point. Grantees also get access to my own custom fleet of agents and an ongoing collaboration with me. The goal? To close the gap between an idea and the world.
REID:
So, Jonathan, thank you for joining us on Reid Riffs, part of this token grant program, which I love—meeting people who are enormously creative, who are leading the edge of what kinds of things to do. And so with that, Parth, why don’t you kick us off with an introduction to Jonathan and how you added him to our fellowship of token grantees.
PARTH:
The fellowship, yes. Dude, Jonathan, super glad to have you both in the program and on today to share your creative wisdom. I met Jonathan almost—I want to say maybe eight or nine, maybe ten years ago now. We were working at a startup in West Los Angeles, and Jonathan was a designer there. But more importantly, a couple of us at this company started playing Dungeons & Dragons together after work once a week. We’d take a meeting room and play D&D, and Jonathan, being a design superstar—
REID:
And that’s why I said fellowship.
PARTH:
Yes, exactly. Into the dungeon, right? He had his little iPad and he’d be drawing our characters throughout the adventures as they changed—the experiences, the different characters we had. My character was a purple Dragonborn. And we played for years, actually. It was one of the most interesting campaigns, and getting to know his creative side and the world-building behind him—it was just so much more than being a designer. So really excited to have you on to talk about your creative vision, your ambition, and what you think is possible now that we have some of the most interesting tools available.
JONATHAN:
Yeah, I mean—thank you. It was super fun. Honestly, I’d recommend D&D to anybody. It made me better at almost everything in life, I feel like. So yeah, it was very formative for me.
PARTH:
So let’s—I want to talk about how you first got into playing with AI and what your stance on it was. And I guess, who you were before AI, and what your lens on your creative aspirations was, and how that has changed in the last couple of years since the generative AI moment kind of happened.
JONATHAN:
I happened upon AI by accident. I actually remember when it was—it was like November 2019. I had just started running my own Dungeons & Dragons campaign, so I was transitioning from a player to a dungeon master. And I needed backdrops, landscapes for the game. We had sessions every week, and I’m not that fast of an artist—it takes me like 10 hours to do one piece. So I was struggling, and somehow I found this thing called Artbreeder. Artbreeder, I would say, lets you smash two images together. I know it sounds weird, but you’d pick two images and it would combine them to make something new.
And you had all these sliders where you could adjust the parameters and weights of things. I didn’t know this was machine learning or AI at the time—I just thought it was a fun tool. I was making all these backgrounds, and they were low-res, very abstract, because this was early days of image generation. But the cool thing was, I could paint over them. So I’d paint castles and bridges and things into the image in Procreate. It allowed me to make these landscapes of this world super fast, and the players loved it. They were kind of amazed by this tool, and we had no idea what we were doing.
But yeah, that was my first foray. It was like December, November 2019. And then, of course, things progressed. We got Midjourney at some point. And then I was heavily into Stable Diffusion when it first came out—from 2.0 to like SDXL. And this is going to sound wild, but I had generated like 40,000 images in Stable Diffusion at some point. And the reason was, I was trying to make my own RPG. I had designed the book—my background is in product design; I have a design degree. When I laid out the book, I realized I needed like 45 pieces of art and I didn’t have it.
And so I turned to Stable Diffusion. This was back when it couldn’t do faces or hands that well. But I could draw faces and hands, so I’d draw over those and make the characters. It was really fun—like working with another artist, someone who had these wild ideas. I ended up making this whole book with art from Stable Diffusion that I had drawn over. That’s kind of how it started to evolve. And of course, now there are many, many tools at our disposal—not just ImageGen, by the way, Parth.
REID:
Part of the whole D&D background, which—Jonathan, you may not know this—I started playing D&D when I was nine. So, long time ago. But it’s not just the images and the world-building and the creativity, all of which is super important—it’s also narrative, story, interactiveness, and other kinds of things. So say a little bit more about how the choose-your-own-adventure, make-your-own-world of D&D is not just all the images, Midjourney, et cetera, but also the narrative story, the land of imagination. Amplify that a little bit too because I think, by the way, among other things, it’ll go with a hat.
JONATHAN:
Yeah, so character-building, world-building, stuff like that. For me, it all starts with an idea. Some people are very creative and have all these creative skills—you might have these ideas in your head, and what the AI can help you do is put them together. You just have to know where to point it. So I’d use it as a tool—like, “Hey, I need a hundred names for a magic staff that does X.”
REID:
Right.
JONATHAN:
And then you might tell it, “Make them esoteric,” or “Relate them to this poem,” and get some interesting ideas. Out of those hundred names, none of them might be perfect, but it’s enough for you to pull maybe two names together—and now that’s the best name for the staff. That’s kind of how I’d use it as I’m building worlds or narratives. It’s almost like you’re sitting in a room with someone, bouncing ideas off of them. That’s how I use it.
REID:
Yeah. So let’s go, because this is great. The D&D stuff is the beginning, but let’s move this to a film or creative project. It’s almost like the film is the dungeon we’re going into. So what parts of this feel genuinely creative? Which parts feel more mechanical? And what are the different jobs around this?
JONATHAN:
I didn’t make up this theory, but a good friend of mine did—he’s also a genius and he’s deeply into AI. We were having a conversation about building products—about UI/UX and engineering, but also creative fields like making music, making movies. And the one thing he said: there are cogs and sparks. His background is engineering and QA, and he was like, “I think UI/UX and music-making and these creative fields are sparks.” That’s where the power is, that’s where the human comes in, and that’s where the AI can’t really go that far. The cog jobs are more logic-based things.
So for my project, I have the vision and I’m setting everything up. I don’t want to offload things like storyboarding to the AI. Of course you could—you can ask GPT, “Storyboard me a movie about a dragon,” and it will. But—and this might be the wrong theory—I feel like with those kinds of things, it regresses to the mean a little bit when you rely on that creative output. The AI is giving you the best average, right? And true art, creativity, is improbable—it’s statistically not probable. So storyboarding, for me, is something I want to do. If you relate it back to, say, Kurosawa—
Let’s look at a movie like High and Low. The way he frames a scene is based on composition. He might have someone in the foreground, someone in the background, someone in the middle ground, and then there’s negative space right here—and he’s getting you to look at that negative space. For me, I don’t even know how I’d prompt that to get the perfect composition. That’s why I want to do the storyboards. That’s a spark job to me—that framing is really hard to achieve without the human creativity part of it. So yeah, that’s my theory behind it.
So I put things in categories: spark jobs and cog jobs. In the short film I’m making, the cog job is—there’s a very mechanical part of this film, and that is, I’m trying to make something that looks like 1980s Filmation. Hanna-Barbera, Ralph Bakshi—the first Lord of the Rings animated movie, where they’re rotoscoping over a person, over a scene. To me, that’s a cog job. Once you have the style, no artist really wants to rotoscope, because for a five-minute movie you’re going to be drawing over like 7,000 frames. That’s not fun, because you already have the costumes, you already have the aesthetic. So that’s something I offload to Luma. Luma AI.
That way, Luma’s doing the rotoscoping and I don’t have to draw 7,000 frames—because I designed the characters, I designed the scene, and I shot the scene. But Luma is doing the cog-job part of it, which is the rotoscoping.
PARTH:
So you really think about this more like AI acts as a studio that you’re directing on your behalf, and you’re taking the high-level—you’re the director. What do you think you fit in as, if you view these tools as doing the execution, the cog jobs?
JONATHAN:
Another example is the song I use in the opening of the film. I had played that on guitar 10 years ago—I recorded it 10 years ago. I remembered I had it, and I thought, this is perfect for this film. So I put it into Suno and made it electric, with drums and stuff, and it fit perfectly. For me it’s almost more than a director—you’re almost this all-encompassing creative. I’m not comparing myself to da Vinci, but imagine what a da Vinci would do with all these tools, right?
It’s kind of like that, where I think the people who have a lot of creative skills right now are going to be able to almost make anything in their mind. We’re getting to that point. So I don’t know what I’d call myself—I guess creative director, a little bit of everything.
PARTH:
It’s interesting. You get to put yourself in whatever level of it you’d like to.
JONATHAN:
Yeah. If you have a breadth of skills, it’s like you’re picking your spots—where you’re best and where your creative skills are. I think that’s a huge advantage. There are things I can do now, like making this short film, that before I would have just stopped at the storyboards or the script. I wouldn’t have known how I could actually make it—without AI video, I don’t think it would be possible.
REID:
Well, let’s amplify this. Go big. And not just because we’re talking about film—we’ll go beyond 11, for the Spinal Tap friends of our listeners.
JONATHAN:
Oh, nice.
REID:
Yes, exactly. What do you think a five-person-ish, AI-native studio could accomplish—what might have previously required hundreds?
JONATHAN:
That’s exactly what I’m trying to get to. Honestly, number one, I feel like you can make any sort of animation you want nowadays—that’s without question. We’re just getting to the point where we have new Seedance, new Kling, all this stuff coming out that’s really bleeding edge. I think you can compete with actual studios at some point. You have to keep the creative spark within the human part of it, but you have all the tools. With a five-person team, honestly, I could probably make any animation possible. There’s not a question in my mind these days about that.
If you had asked me this question four months ago, I would have said maybe. But we’re making these leaps right now where I don’t have a doubt that it couldn’t be made. Like you said, once you needed a team of 100 animators, or an entire team to create the music, or even to do voice acting.
PARTH:
Right.
JONATHAN:
In my film, we still do the acting. It’s like a mocapped film, right—with Andy Serkis or something, or if you’re into video games, like how they made Death Stranding, where they shot everything with the actors and they’re doing the performance. But the greatest thing is, you can change everyone’s voices now based off that performance. You can change how they look. It’s still based off their performance. So the sky’s the limit. You just need a small troupe—almost like a traveling troupe of creators—and you could make any genre, with any characters. As long as you have that human creative spark, that curation, that intentionality behind the work—
Yeah, you can actually make anything.
PARTH:
Yeah. You kind of already got into this, so tell me more about this project. This sci-fi project you’ve had in your mind—you’ve had this idea for years. You would tell me about it, but it wasn’t really possible until now. What tools are you using? How are you thinking about it? And maybe in just the last four months, how has that become crystallized as possible?
JONATHAN:
Sure. Yeah. I would come up with these ideas, write them down, make a script for them. I’d even storyboard stuff out. And usually that’s where it stopped, because I can’t get props or costumes or actors—or even cameras. To make it would be almost impossible. I have notebooks full of these ideas, by the way, over the years. I would have stopped there and never made these things. Maybe you can pitch it to a studio, but it’s a super long shot that anyone would ever give you money to make anything. I had the storyboard, I had the script, I had made the treatment. Went out, shot some scenes in situ with some friends of mine.
We wore costumes that vaguely resemble it—but again, they’re going to change in Luma. The way the process worked: I had all the setup already, and then I’d go out, shoot the scenes, and build the characters in Luma. There’s a lot of pre-setup to get the consistency. But that’s something you would have done if you were working at a studio anyway—you’re going to need to get that character consistency. So you build all the characters in Luma. You can even build them in Higgsfield and other things, but I would build them in Luma and then use that as the reference to build the video.
So my process is slightly weird in that—when you watch those old films, like The Lord of the Rings or Fire and Ice, the characters are hand-drawn, rotoscoped, but the backgrounds are painted. It’s a distinct feel. So my process is weird in that I take a reference photo for the scene, and then I’ll change that to a painted style. Then once we do the mocap and change the way the character looks, you can also adjust the background to the reference photo to make it look painted—so you get that feel, that sort of ‘80s hand-drawn-versus-hand-painted-background feel. And that was impossible even last year.
It really was. So that’s how it changed. The way I look at things when I come up with ideas is almost like, “What’s stopping me from making this?”—as opposed to previously, when I would just stop at the character design and storyboard.
PARTH:
Yeah, that’s incredible. Maybe a follow-up to that: what part of creating this film, creating these projects, takes a lot more effort than people might assume? Because it’s not just one click, right?
JONATHAN:
That’s a great question. It’s the human curation part of it—that takes a long time, which people don’t understand. There’s a difference between kitsch and art, and that difference is genning 200 images to find the one, or even 300, or genning a scene over and over again. It’s the curation part, in order to get things perfect. People underestimate how long that takes, but to me that’s going to divide the slop from the non-slop. It’s that craftsmanship that takes a long time. I’d also warn that the character setup takes a while to get that consistency.
It’s very important. You need to go into it and fully design these characters—down to what jacket they’re wearing, what pants, what their boots look like, what their gloves look like. I’m making a weird fantasy western, and the characters have a very distinct style. It’s almost like in an anime, where the main character wears red and the side character wears blue—almost everyone has their own palette. That takes a lot of setup in the beginning, too. So that, and the curation. But again, it’s about the intentionality behind the work. You’re not just pressing go and genning.
And the first gen is rarely perfect. You might have to go back—and the cool thing is, I could reshoot in my backyard. I could go back and reshoot a scene in my backyard, make sure the framing is correct, and then do that one. It’s that part, I think, that takes the longest.
REID:
I think you’re right that curation is a very key word here. There’s this question around whether curation is all human—individual humans, groups of humans—or all machine. But more likely it’s a combination. The notion of how this pulls together in terms of human and machine collaborating, dancing together, is really important. How has your perspective on curation changed in this new AI-amplified world—where you have multiple humans, multiple agents? And what does that mean for how we get to what is meaningful creative work?
JONATHAN:
My perspective has changed a little bit, in that I think we have access to a lot more. Before AI, you were a little limited in what you could explore. I could only explore things I was good at illustrating, or things I had on hand—maybe you could search the internet, sure, but it was very limited. Now you can get wild with production. It’s one of the reasons I kind of prefer Midjourney for exploring, because Midjourney still has a little bit of this artistic wildness to it, where some of the other things, like Nano Banana, are very pointed and distinct.
So I like the randomness. I would almost rather gen a hundred random things to find one good piece out of it. So I think our access to ideas is completely amplified, and that’s where the curation is evolving—now you have access to so many more ideas than you would have previously. It’s almost like giving you new directions. One thing I loved about the early Stable Diffusion days was that the image gens were so wild, it was almost not even what you were prompting—there was something chaotic about it that I loved. It’s just like an idea generator.
PARTH:
It reminds me, back to D&D—it feels like chaos magic sometimes. It’s not random—it’s actually not random; it’s a probability distribution—but it’s also potentially something you’ve never seen, an idea no one’s ever combined and remixed. And then you can just go to latent space and explore the space of possible ideas to find that one thing no one’s ever seen, that might need to be seen by someone. Might be worth showing to the world.
JONATHAN:
Okay, I’ll give you an example. When I first set out to do this project, I wanted that rotoscope feel for everything. When I started genning in Midjourney, there was one out of hundreds of images—one from my references was slightly different. It was a little more animated, a little more exaggerated, and it just popped out to me. So I started making all the other ones like that one. And so it changed from being this completely rotoscoped, old-school style to a little more—almost like anime, but a cross between the two. That’s something I would have never found on my own. I would have never stumbled on that.
Even if I was drawing it myself, I would have never gone down that avenue. And so that’s, to me, where that sort of curation lies—that interaction between the amount we can produce and the access to these ideas. That combo—I would have never found that without genning those. So, yeah.
PARTH:
Oh man, that’s so exciting. Well, Jonathan, I’m so glad we could have you on. I’m so grateful to have you on this journey with us, and to see what you’re creating. Thanks for joining us today. But before we leave, I’m wondering if you have a question you’d like to pose to either Reid or myself.
JONATHAN:
Sure, yeah. I do have a question. There was a part in your book Superagency where it says that true super agency is defined by allowing humans to tackle higher-order challenges. So my question is: are there some critical higher-order challenges that you believe we are currently not addressing—and what are they?
REID:
Great question. There’s a set of different things. Part of the writing of Superagency was to try to show people that, even though they feel their agency is diminishing because the world is changing around them—technology is being developed that changes the landscape—they can still grasp agency. They can still do things. Even though, for example, Apple builds an iPhone, which you may think reduces your agency, you can use the iPhone and gain a lot of agency. That’s the parallel for the call to agency that Superagency is. Now, I do think there are some really important, more subtle questions here.
So, for example, even though there’s this technological tidal wave coming in—and you can’t choose the tidal wave, you can’t choose the timing of it—there are still ways of gaining agency. But how do you gain it? Partially, not just by individually assigning tools, but collectively. What does this change in terms of our landscape, the social environment we’re in? Part of the reason I use the iPhone as a parallel is—well, sure, it enables you: you can take pictures and have maps everywhere you’re going. But of course, it also has everyone else taking pictures. What does that mean in terms of the universe you’re now playing in?
And what does this mean for our collective knowledge, our collective epistemology, how we parse the world? Some of that gets to: how do we discern truths about ourselves, truths about the world we’re in, collectively, together? How does that play into how we gain agency, not just individually but together? One of my hopes for AI is this notion that, if we’ve got a good sense of how it amplifies our knowledge of ourselves and the world—what is true—that then helps us learn collectively better. AI is one of the best learning tools developed in human history to date, because it builds upon books, it builds upon the printing press, it builds upon the internet.
And I think that sense of collective learning—learning together and learning truth, and what is truth, and the epistemology of it, not just outsourcing it to AI but us learning together—is one of the key elements of that kind of collective agency.
JONATHAN:
Yeah. That’s kind of part of the book too, that resonated with me a little bit. Is that the bloomer ethos?
REID:
Yes.
JONATHAN:
Yeah. So, definitely—I’m definitely part gloomer in the book, but I’m also a pessimist about a lot of things. So I have a doomer aspect to some of the things I think about. But we’re all conflicted, because I’m using these tools every day and creating wild things with them. So when I would read the description of the bloomer, I’m like, well, that’s definitely what I’m doing.
REID:
I actually completely agree—on Superagency, there’s kind of bloomer, doomer, gloomer combinations. It’s a more complex thing. So thank you for joining us, and I look forward to seeing some of your creative outputs in the future.
JONATHAN:
It’s going to be awesome. I will 100% send you what I finish.
REID:
Possible is produced by Palette Media. It’s hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Shaun Young. Possible is produced by Thanasi Dilos, Katie Sanders, Spencer Strasmore, Yimu Xiu, Aman Suri, Danny Garrison, Trent Barboza, and Tafadzwa Nemarundwe.
ARIA:
Special thanks to Surya Yalamanchili, Saida Sapieva, Ian Alas, Greg Beato, Parth Patil, and Ben Relles.