What Future with Joshua Topolsky[rerelease] What you’re reading right now was written by an artificially intelligent — though not sentient — neural network designed to write descriptions for podcasts. Or was it? You don’t know, and that’s what makes AI so fascinating, cool, and scary. Since Josh loves all of those feelings, he sat down with the founder of controversial AI-generated art startup Midjourney, David Holz, to unpack the future of creativity. Discussed: Porn, therapy, families, Hitler. This one’s a doozy!
Today's episode of What Future with Joshua Topolski is a re release. It's a really interesting listen. We will be back next week with an all new episode.
Thanks for listening, hey, and welcome to What Future. I'm your host, Joshua Topolski. Now, I don't know if you have been following any of this stuff with this new artificial intelligence machine learning driven AI driven bots, So there's several bots that do this. One has recently become widely available to the public. It's a bot called mid Journey. From a user perspective, From a person's perspective, what you can do is you can sit down and type something and it will generate based on all of the images it's ever looked at, which are billions of images on the Internet and wherever else. Databases they feed it and stuff like that. It will generate what it thinks you want from a prompt from a sentence. You actually use it through Discord, which is like kind of a chat you know network, which was popularized by gamers. But Discord you can basically go and talk to the discord bought from mid Journey, and you can say, for instance, Dracula explaining his vamporism, to a crowd of onlookers, and it will generate four different images that it thinks capture your idea, and they're insanely accurate. They look like somebody painted a picture of something that you wanted. It is essentially like the closest thing to being able to visualize a dream. I don't know that I've ever had my mind more blown by anything a computer has ever done than this piece of software. I mean, it is hard to articulate what it feels like when you write a sentence of something that seems completely impossible and then see a pretty good representation of it in a matter of like thirty seconds, forty five seconds, maybe a minute, to see your first four versions of something. Let's just say you're not an artist, You're not a designer. You're not going to make your living off of doing paintings for magazines or whatever. You're just a person from a pure, like thrill level as a person, I think this is fucking amazing. It is like the most fascinating and most amazing thing I've almost maybe I've ever done on a computer. And I've done a lot of stuff on the computer, you know what I mean, But this is like trippy I mean it's fucking insane. It's like, like I said, it's sort of like as close as you might get to you know, you had a dream, and you can then see the dream. I would also say, what's interesting is that it is like a computer dreaming. I mean, what it is is you giving This is like what I consider to be fairly abstry input to a computer and the computer deciding making all of these really creative decisions about what that thing should look like. Anyhow, I'm not an artist. I'm not a painter. I have been working with mid Journey to create art for this podcast, and you can see some of the prompts for these, some of the ones that I did science fiction paperback book cover about society in the future. One of them is a phrase from Blade Runner attack chips on fire off the shoulder of Oryon, which is Rucker. Howard has this monologue at the end of Blade Runner, and that's one of the things he says, and it created imagery based on that sentence, Dracula explaining his vamporism to a crowd of onlookers. You can see several variations. I mean, they are fucking beautiful pieces of art. In my opinion, like legitimately beautiful pieces of art. There is some art in figuring out how to get this thing to do what you want, or to at least create a result. Is that is pleasing? Now? Anyhow, what's interesting about this? There's many interesting things about it, and like I'll just go down the list of some of the ones that I'm thinking about. First off, there's obviously this question about art, like what is art and is this art? And what kind of art is it? Meaning As a guy who who's run a lot of newsrooms and a lot of publications, I could see this is very functionally important in like an organization that needs original art for things, but maybe it doesn't have the budget or the time to generate original art for everything they'd like to generate original art for right, So there's an implication there, like for me that I'm like, oh, that's really interesting, Right, that's really exciting. I follow a bunch of designers and artists on Instagram and they've been talking about this for a while. I mean, this opens up an enormous amount of serious questions, like, for instance, the bots are obviously taking content and material and analyzing it and learning from it, and in some cases, replicating it in some way from a real artists right from historic you know, pieces of art up to modern pieces of art. As far as I know, in essence, these ais can go and look at and then learn from. But there's this little bit of controversy, or not a little bit, maybe a lot from some artists who say this is you know, it's theft of our work. You know, they're using things we've created without any license to do so and creating new works based on it. That argument, to me is a little bit like every artist uses somebody else's work to create what they do. I mean, as we know, remixing in music has become one of the baseline ways you make music now, no pun intended on baseline. So the idea of like sampling somebody else's art to create something new is not new. I think what's sort of insane and threatening on a bunch of different levels is that this is creating real art, really interesting pieces of art and imagery that have real applications, whether it's hanging in a gallery or using for an illustration in a magazine or whatever, and it is just removing a person completely. Basically, this image that I created for the podcast is a perfect example, I could sit with a designer and tell them about what I wanted and show them examples can you make something like this, and we could work through it over and over again until we got to something that fell right. The idea that I could just say it and it could be like, this could be it. This could be the art for the show. And that's one job that a artist is not going to get now, like for sure, right, there's an implication for people who work in these fields that is way different than what we're talking about. Like I was like, at a pure human level, this is thrilling, but on the flip side of that, there's entire industries that potentially are wiped out by this. What does this open up, I think is a question that I don't know the answer to, which is, in five or ten years, this is going to be so much more capable to create things like this, capable to a point where I think it's likely in the next five to ten years you can simply tell it to do something whatever it is, and it will create a perfectly photorealistic version of it. I mean, and there are versions of this where you can say, you know, show me this thing and show it to me in these different styles, and it'll show you an image in the style of this painter, or or like it was a photo taken from this era, or like it was you know, shot on a certain kind of film. What that means going forward is almost like kind of frightening. Like people are like talked a lot about deep fakes, you know, and they're like, oh yeah, like they're going to fake a voice, or they're going to fake a person's face or whatever. Like this is essentially like we're getting to the point where you can just fake any situation. You can just create visually any situation you can think of. And I think the logical thing is that eventually, pretty soon, I would imagine it'll be able to do this with video, right, and I think with moving images sound is not too far behind it. You start to think of how this could be applied to all sorts of other things. I mean, presumably, if it can do this with art, with visual art, I think it can do it with other forms of art. Right. Will we discover when it's like you could have any art available to you, any type of content available to you, perhaps that like what you want is somebody else's brain in mind, Right, like I want to like understand or see something or hear something from somebody else's brain, but I don't know what it's like if if the other brain can just create any of those things that I would be intrigued by, Like this art is a great example. Clearly, this non brain entity can create things that surprise and delight me, that feel as authentic and original as any art that I've looked at. It's obvious that the systems that are creating this are very advanced, and they are only going to get better. They're not going to get worse. There is no going back to a state where this is not possible. And so when you think about what that looks like down the road, like maybe not everybody feels this way, but I can kind of like in the middle of my brain, I get this like very upsetting feeling when I think about what space actually is, which is like this endless nothing and actually nothing and what is that like? It's very upsetting to think about to me. When I think about like the future of this stuff, it's sort of a similar kind of weight in the middle of my brain, which is like where does this go? Like it feels like all of real is almost called into question by the technology, and maybe I'm overstating it, maybe I sound crazy. I'm not saying the computer sension, or it's alive, or it's got a soul now or anything. But there's something in between the lines of all this where it's just sort of like it leaps beyond even my understanding of what is happening, Like it leaps to a place that's almost like, I don't want to say spiritual, but it leaps to a kind of almost religious place where it's like, how can this be? You kind of feel like when you do it, how can this be?
Like?
How is it possible? My guest today is David Holds, the founder and CEO of mid Journey. David, thank you for being here.
Thank you.
Just before this, I said, can I say CEO? And you didn't want me to, but I've done it anyway. We're all going to have to live with the repercussions.
I no, I've been exposed.
Okay, Let's say you and I met at a party. Let's pretend we're at a cool party. You don't know where I'm coming from, and I'm like, what do you do? And you say, I'm the founder CEO of mid Journey and I go, what's that? How would you describe it to somebody just randomly at a party.
I try not to. I'm pretty low key, but.
If they ask, they're like, what does mid Journey do?
Yeah? I I never really wanted a company. I just kind of wanted a home, and so like mid Journey is sort of meant to be like my new home for the next ten years, to work on a lot of cool projects that I care about, with with cool people, and that hopefully are are good for everybody else too. I know we have sort of themes that I want to work on, but I had to put in three words. It's like reflection, imagination, and coordination. Like in order to flourish as a civilization, we're gonna have to like make a lot of new things. And making new things involves those three words. Wow, and we need a lot more around them, like infrastructure, new fundamental forms of infrastructure really around each of them. We were actually originally working more on the reflections and coordination tools. We were doing some imagination stuff, but then there were certain breakthroughs on the AI side that were happening. It was about like a year and a half ago. Now it looks like everything's blowing up. But like a year and a half ago in San Francisco, we all went to the same Christmas parties and stuff. All the AI people are kind of out in here, and we were kind of all together, and I'm like, these diffusion models, it seems different than the other stuff.
And they're like, yeah, no, this is different, and well, what are you gonna do?
What are you gonna do?
We're all kind of just talking, and I'm eventually like, I think there's going to be a human side of this, that it's not just about making pictures, but that there's a sort of a back and forth. There's like a lot more to this that's gonna be hard to figure out from just optimizing a single number and up computer program. Right.
There may be some taste involved, and no one knows what that is. And I'm like, i think there's something I have to contribute to, right, Yeah.
Can you imagine? Though I'm a guy you just met at a party. I've got no context whatsoever about mid jourdy and you just told me that, which, all, by the way, all very interesting. I have many questions related to what you just said. Yeah, I had dumb it down a little bit only because maybe not every single person will know, but ah, mid Journey is known right now. The company has risen to kind of a place in the spotlight because it is what I think we're all sort of talking about now, is like an AI art tool or a tool to create art based on artificial intelligence and machine learning and all of these sort of other very complex technologies that are kind of fusing together to make something that is relatively new. So I think most people would say, you've built a tool that can take human language tax like basic like English prompts or whatever. Maybe you do in different languages, I don't know, and convert a prompt like a description of something into a piece of art that is created basically wholly by a machine. Is that correct?
Yeah? I try to avoid like the word art almost to be honest, okay, because I think that it's like not really about art. It's about imagination and sometimes people use their imaginations for art, but usually not right, And so I usually think of it as we're trying to create these machine augment imaginative powers. Sometimes I almost call it like a vehicle, right, you know, to really like to ask like what are we doing? Like is it like the invention of photography and how it changed painting, right, And I tend to say no, it's much more like the invention of the combustion engine and the car. And like when we invented cars, they're faster than us, but we didn't chop our legs off.
We don't have to really move somewhere.
You move through vehicles, So it's kind of like a vehicle for a magines If you really have to go somewhere, you're going to unit these vehicles like jets and boath and cars.
We never have a little robot as like our icon.
It's like a sailboat, right, you know, very much trying to kind of help people explore and imagine these like seeds of like esthetic possibilities.
I mean, it's interesting that there's a little bit of like a defensive stance you have to take now because the art aspect of it gets under the skin of a certain part of the audience that's like, wait a second, you know, what is this thing doing? What does it mean? What does it mean for all these different industries. Yeah, I think a lot of people feel and maybe you guys have had to play some a new round of defense because of it. That this is an engineer to kind of like up end industries, right, But you're saying you don't really view it that.
Way, No, And that to me is actually very uninteresting. Like the idea of like making fake art is really uninteresting, like who cares? Or making fake photos it's really like to me is not like I think, what's interesting is making stuff that never could have existed before. I don't like it when somebody makes a deep fake photo of the dog, we make it really ready to do that.
Other ones do that well.
To me, the most interesting images are the ones that don't look like anything we've ever seen before. They don't look human, they don't look like the AI made. They look like something new, and all we know is that it's this new thing, it's this new frontier.
Right. I should tell you that the art for this podcast, as of right now, and I don't think it's going to change, is generated by mid Journey Cool, and it ended up producing results that I think are like at once very familiar to me, like in terms of stylistically, there's something very familiar to me about it, but there's also something about it that is like totally original, I think to your point, now, I'll tell you this, like, I'll give you my stance a little bit, because one of the reasons I wanted to talk to you, one of the reasons I want to talk about this at all is, as you know, I'm a huge nerd and I've spent my entire life like being, you know, sort of mesmerized and interested in emerging technology in all sorts of different forms. And when I started using mid Journey, mid Journey's producing something that to me feels I'll try to avoid using the term arte, it feels like it's creating something very original. I could say, like, Okay, I know where some of this stuff is coming from, Like I can kind of understand, like there's certain styles that are present, or if you give it a prompt to get a certain style, you can get that. But to me, it was like, and I still feel this having processed it now for you know, weeks and months, maybe the most amazing thing that I've ever seen a machine do. I totally understand the idea that you're not trying to build a tool that is like a new photoshop, although I think or applications that are obvious that are in that realm. When I first asked about what it was you use three words. What were the three words that you used.
Was reflection, imagination, and coordination.
Okay, so coordination and reflection. I want to talk about like what that means because I understand the imagination part, and I think I understand how you are thinking about like what mid Journey does now in that department. But tell me about like those roots of reflection and coordination, Like what was this before it was? What it is?
We were working on a lot of things trying to like understand human minds, like individually to help people reflect and then also to kind of help people come together and like work on things better. And so we were doing a lot of like quantitative psychology and like structured thinking to kind of like create like boost up of a hive mind as fast as.
You can kind of a future is going to say lots of weird things.
No, that's good. Are you saying that? Like the roots of this are kind of like can we get this thing to think on a collective level for us to like solve problems? Yeah?
I think there's two areas.
There's both like how do you help somebody think about like who they are, what they want and just kind of like deal with their things. Uh. And then there's also like how do you help them find like the right people anything big nets other people, So how do you kind of find the people?
And I don't know.
When I was like twenty, I would say, you have to have your goals and then you align people who share the goals.
And then I've done that.
And it turns out that the second the goals change, the groups blow apart because like it's about values or something. And then if you a lot of people on values and then over like five or ten years, it blows apart again because it turns out that our values like change in our lives, our experience change, right, And so then maybe this idea is like we need some higher than values, and maybe it's aesthetics. It's like not about what's like right or wrong or what's important on importance. It's like really deep down, it's about like what we feel is beautiful and what we feel is ugly that like really leads to the things that we value, the things.
That we actually tried to build.
It's interesting.
And so there's this idea of like maybe aesthetics themselves are like some of the highest things, and maybe aesthetics can be like a foundational layer of like a social world in a way that like is beyond where it is, because right now it's like the Internet, what is it?
It's about?
Like Facebook, it's like who's your mom and who went to school with? And then like on Twitter, it's like almost like like you say one thing a day that pisses people off and then half of them will follow you, right, And like those are both shitty foundations for like a better social world. I would never want to build a team that way, so that there's something really interesting on like mid journey where people come together and they're like, man, you love like Egyptian space pyramids too, That's like me, And then like you have nothing else in common, but you both love Egyptian space pyramids and it actually like means something really deep, right. I think that like aesthetics have the potential to be a foundation of a better social and cordant player in a way that's like really hard to understand, but that is actually like really interesting.
I mean that's a fascinating and frankly I have so many questions around just the basic concept there, but like I would agree with you that aesthetics do tend to bring people together. I mean, but aesthetics conceptually, the idea of, you know, having a taste or a preference for something, there's a limit I would imagine too, people who identify around an aesthetic position. Meaning my mother, who's a wonderful, wonderful and extremely insane person. She could talk about things she's visually finds beautiful or whatever, but I would not say it's like a central part of her personality or something that she has an enormous amount of interest in. Right. The thing about Facebook is that a raw opinion or sharing something like oh, I found this article interesting or whatever is very straightforward in the sense of we all know what an idea is or an interesting article or an opinion. But I don't know that everybody thinks on an aesthetic level. Maybe I'm not giving everybody enough credit. It's possible. I think you're right.
People don't think about it, but it's there, right, Like I tried this, I'd like, what are your aesthetics that lead to your values, that lead to your goals? Like you can ask the question and almost nobody can answer it. Right, It's a really hard question, But all of a sudden you give them something like mid journey and it's like you can make a picture of anything, what do you want, and like everything just spills out and then they go through this whole like heroes and mid journey and like the process of looking through that journey, like you like it's all there and it's very clear, like a lot of stuff comes out.
Actually, But if I'm like a and I'm gonna give you like a really extreme example, and so forgive me if this feels like a like a a gotcha or whatever. But if I'm like a neo Nazi, for instance, yeah, I might love Star Wars. Let's say, although I always find it fascinating when like people who are really into fascism, like are like I like Star Wars, I'm into the Rebels or whatever. I'm like, you know, it's but okay, let's say I like Star Wars. You like Star Wars, but like one of us is a white supremacist and one of us isn't. We may share some aesthetic interests, right, or we may both love a certain artist, right, you know, we're Lichtenstein fans or whatever. But like, at the end of the day, deep down, I don't know that that aesthetic preference has any deeper resonance on who we are. There's a limit, right, well, so.
Like for example, like you know, you're a rebel obviously, and then like a Nazi nowadays is also rebel in their own way, so you do have something in common. Yeah, but like, but there's probably also other things. I mean, that's that is a leap, I would say. I mean, I get what you're saying. They're definitely going against the grain, right, I got it. Yeah, I get the grain. Yeah, I know, I mean they were Yeah, definitely a lot of us are rebels, and they're types of rebels, but we are rebels. But now, like I think there are other things too, right, So you don't want to just lock on rebels. You want to have something it's like a little bit broader and more interesting. It's the question is after you make a bunch of picture of rebels, what's the next thing you do?
You know, and then what's that's? What's that all come together?
You know? I mean, now we're like very far afield from Like I've got a mid journey bought that I can talk to and it can make images for me. I mean, how would you describe it? You describe it as ai?
Yeah, I mean it is it is ai. I don't like.
I kind of avoid the words ai and art actually both together. Weirdly, problem with words like AI is that people give it a lot of agency and like will and purpose and meaning right and so where it's like this thing it doesn't have a story or a narrative or like any will.
Right, ascension doesn't have a soul.
It does learn actually from lots of people, and it changes and there's a co evolution. It's almost like mid Journey is a flower and then the users are bees, and like the flower is trying to be beautiful for the bees, but the bees pick which flowers are the ones that get to survive, and so like there's this coevolution between the flowers and the bees.
Like there's not a lot of will. There is some will, there's a will to be beautiful, and then there's only weird about flowers being beautiful because we find them beautiful too. It's like, what does that mean? It's because they're not really for us specifically. It's like why do both us and the bees find something beautiful? Like it's sort of speaking some weird objective thing.
No, And I can understand that from a philosophical level. I mean, like, what is it doing.
It's a program, Yeah, it's a program. It's a program with a lot of models in it. There's a model that models language, and there's something that models the connection between language and images. There's other thing that tries to model what images look like.
Right.
There's actually also models that try to understand like beauty, like what is beautiful actually? And then there's other models that try to understand like trade offs between like diversity versus, creativity versus like how literal should you be?
How metaphorical should you be? How do you read things? And so it's kind of a it's.
Like a structure and there's a lot of like ducta and you know, it's it's weird because like people will be like is it alive? Like, well, how is it understand thing? If I say something like sadness or happiness? How is it able to make an image of an emotion that it's never had? Like they go ask these questions like what is this like that's it doesn't not like a piece of software, you know, but it's not an AI because it's never had those experiences, right, Like what does it mean?
There's a lot of really interesting questions.
I think a lot of people they hear AI they think there's like a machine somewhere with like a glowing red orb in the middle of it, and like it's like pulsing. Yeah, exactly, And there's like some neural net. You've built some custom hardware where there's like there is the neural net. It's like a digital brain. It's like software. Right.
These programs they do share things with our brains, like like how an airplane share something with a bird, Like they both share aerodynamics and physics and the sky. Like these things are sharing some physics of thoughts.
Right with us. Right. But I'm just saying, like, it's not how you built software. The software does some pretty sophisticated things. It is hosted on like a AWS rack somewhere essentially, I mean maybe now use AWS. So what is the product? Like, you've got investors, right, No, you don't know your boots dropped? Yeah, okay, now I listen. I've paid for a subscription. I'm a mid journey free now, So is that the product people pay for subscriptions to use it? Yeah?
I try to have very honest business.
It's like you're not going to run on your computer ground the cloud and then we're gonna pays takes money and then we'll take some margin on that, and that's the business.
And you feel like that's a good foundation for like whatever this thing is going to be, Like you can build off of that. Yeah, you don't have like Mark Andrews in company being like, I'll give you X number of billions of dollars if you can let me turn this into whatever Mark Andrewson wants.
We do have a lot of vess coming to us offering us lots of money.
You're not taking the money.
No, we haven't taken anything so far.
That's pretty amazing. Can the business be profitable like this?
We're profitable already? You are, Yeah, that's one reason not to take money is we're already profitable.
Well, I mean, if you're making money, it's definitely good reason not to take it, right Yeah.
I mean if people come to us and they offer us money and I'm like, what am I going to spend it on? And they're like it's good to have it, you should have it, And I'm like, where you have money, We're trying to spend it already, and they're like, well, you just have money.
They just take us take our advice.
It's not about the money, it's about advice, or like they try to make those arguments and so far having curt a very compelling argument.
So you're happy to iterate on this product where it's at now and let the user sort of maybe dictate some of the direction because of the way that they're using it.
Yeah, I mean it's kind of beautiful.
It's like we make something and people like it, they pay us money, and then if they don't like it, we don't make money.
But like so we have We're like we're trying to make something.
People like because it supports our stuff, and like it's very sort of honest to straightforward, and it's the easy business.
I keep it this way. I would keep it this way.
I mean, presumably there's commercial applications for this, right, Yeah. I think of this because I'm like a guy who runs like media businesses. Yeah, I think, oh wow, there's all the time, like I want art for something, and I'm actually going to get into a bunch of questions about the art side of it. But all the time, like I'm in a newsroom, I'm like publishing you know, twenty stories a day or fifty stories a day or whatever, and every one of those pieces has some art attached to it, Like presumably you're already doing more enterprise level stuff where like I just want like some design for a story or for a blog that I'm writing or whatever that you could generate that any sort of infinite iterations of original pieces of art, Like, is that a part of the business.
I would say, we're a consumer business that also has like some professionals. So it's by like seventy percent consumers and thirty percent professionals. Right, the professionals are mostly using it for like brainstorming and concepting. Then the consumers are having fun and sort of having these reflective, spiritual personal experiences. I'm not that excited by professional use, even though, like I'm happy when I see people are finding it to be useful. Right, the regular people have definitely been a lot more motivating and inspiring to me, even the professional uses.
I have very little interest in the world as it is.
I want to like make it really different, and it's much easier do thing really different fornsumers than it is to like have that immediately impact the sort of professional worlds, right, and so like for video game people come to me to like well, and like they literally they have to file us under their photoshop budget because like the video game is already budgeted out and it takes sixteen months, and I have to wait for them to make their next video game. And I'm like, this is bullshit, right, I'm so happy that my business isn't reliant on somebody finishing their video game in sixteen months, you know.
And that's what that world.
Is like, right, Listen. I thought a lot about this, Like, if I'm making a video game, especially if I'm like an independent developer, like an indie dev I need art, I need assets, like I want to make like I want to make this world that hasn't been made before, and like normally, and this actually gets into this part of the conversation I want to have about art and about the sort of implications of it. You know, I might go and hire an artist or whatever to do that, but now, like mid Journey, potentially, like if I'm using in that way, I can create assets and backgrounds and scenery or even brainstorm off of that to build something from. Like That's that's not the exact thing, but a kind of iteration of it. But there is a certain very vocal segment of people out there in the world, and there are people who are artists who are you know, digital artists, or who are working artists today, or even people who are doing fine art that's like hanging in galleries. And they're like, one, this is theft because it's using our work. It's using work that is out there that is available to see as inspiration for these works. And two it's like they're not getting anything when it does create new work. Not only is it making their jobs more sort of obsolete, but it's also like doing it on the backs of all of their work. It's not a non compelling argument. There is some reason to think that all of those sort of notions are in some way true. Like what's your take on that.
There's a lot of misunderstandings around the technology, and it makes sense that like artists really aren't going to understand what this is doing. Some of my favorite images I've made with any of these models that looked artistic. We're trained only on photos and so but this is is it's a system that understands what images look like like. If you've seen enough photos in your life and then you see it anything you could describe the painting without having ever having been trained on paintings, and so like, what this is the thing that understands images, and then it understands language of the connection with languages and images, And there's some elements of like knowing what a style looks like requires having seen the word and the style before. So there's like some connections to it, right, But like, largely speaking, it's not I think working like the way they think it is. And so the problem is that the artists are scared about being in the data set. But literally you can just take one of their pictures and feed into one of these models where that ever, never have it seen it before, and it can make pictures like that.
So it's not about the training data. First off.
If it understands images, it's game over for that battle. It wants to seem enough general images enough to like know what textures are and know what colors are. You can show it a picture and it can make pictures all like that, never having seen that specific car asia before.
Right, So I mean you know that obviously raises like all kinds of weird questions about like you know, how fine tuned does that get? Can I pick any artists like a photographer I like on Instagram and say in the style of this Instagram photographer, and like it'll do something well.
I mean, you could certainly put a photo of theirs into another service and I'll give you a photo that looks like it. So you know, that's I think that's really the more that's kind of the technical thing, right, And so basically, if these systems understand images, they'll be able to copy anything you show them, regardless of whether they're trained on them. So I think the training data is the wrong battle to fight, But there is potentially a battle to fight over like use of these tools, like what is good and what is bad use? And certainly, like the law covers out already, if you make something that's really derivative of another artist, like too derivative, it does not okay even legally, right, So like there is it is covered a little bit by law already. Maybe there should be something more strict because like it's getting easier. But that's that's the battle to fight. I think it's like what's too similar, not like this training data thing, you know.
I think about like CGI in a way, if you're building an environment for like a film or something, right, like and you're like I want to make a mountain or whatever. You're not going to hand draw every polygon that builds the mountain, right, the computer is going to figure out and even now, like it'll just basically terraform a mountain right in unreal or whatever.
Yeah.
Yeah, And once upon a time most people couldn't read and write, and now everybody can. And there are more writers now and more readers now professionally than they're ever were before.
Right, Well, it's kind of like photography, yeah, right, Like everybody has like a kind of pro grade camera in their pocket all the time now, and so like we were just a wash in really high quality photos, whereas like if you go back fifty years, not even twenty five years, the best phone camera you could have was really shitty and was obviously low quality. We weren't a wash in just photos everywhere. Right, And like in the last twenty five years, pretty much everybody's become like somewhat of a pro am photographer. Yeah, maybe this is a straw man. I don't want to throw strawma at you, But like, is there a question about like deep fakes and sort of like creating reality that does not exist? Is that something that you guys grapple.
With yeah, I mean it's a real risk for us. Specifically, we did some special algorithms. It's very hard to make it, make it deep fake what it does. If you ask me to make a photo, it'll look realistic, but there's like something to it in the lighting and the shading and the hues where it's like just far enough away from a photo that it looks very realistic, but your body like knows it's not a real image immediately.
What if I'm imagining something that looks exactly like a real image, We're not doing that right now. My imagination has a limit.
Yeah, right now it does. Yeah, yeah.
Would do you think that limit will be lifted for certain users? Maybe for this guy, I guess the very creative ideas. Maybe let me check it out.
There's lots of pros and cons to doing that.
So we found that when we flipped it over that boundary, sometimes it looks perfect, and then sometimes it looks really like Uncanny Valley zombie like right, it's like upsetting. And so right now, if we flip it, it's like kind of say fifty fifty fifty some times looks perfect, sometimes looks like uncanny, and the uncanny is so like upsetting to me, as like a visual aesthetic person, I don't want to make anything that looks like that, right, and so like I just it's better to just not.
Allow that at all.
Maybe in the future it'll be so good that it never looks unkenny and it'll take the technology is not.
Yeah, I mean there's no chance, just to be clear, there's no chance that in like five five years from now that we won't be at a point where mid Journey or other programs like it will be able to create completely photorealistic, if not full moving images for sure, still like in five years time, right.
Yeah, Yeah, There's got to be multiple directions here.
I think one I'll be trying to like make photo realistic duplications of reality, and I think the other want to be like making things that are sort of super real, like beyond real, right, And I think the beyond real stuff is where it's both interesting as a human and probably where all like like consumer and commercial stuff is.
I will say, I'm unabashedly like a fan of this thing, but like I also can understand people's fear about it. But people are afraid of a lot of things that computers do, and for very good reason. I would also say, and this is kind of your problem. People are afraid of people like you. I don't mean you personally, You're a lovely person as far as I know, but like you are like, hey, I am interested in imagination all these things. And like if you ask like a Mark Zuckerberg, like the early stages of Facebook, you know, he would be like, I just want to connect people, you know, I just want people to like get together in this social and environment or whatever and connect. But like, actually down the road, as that thing developed, Mark Zuckerberg made a lot of like really crazy, weird, bad decisions. They don't have to go on record by agreeing with me, but I think in your heart you know it's true. And so what do you do to protect against like these things that feel like creative decisions now? Right, Like, we couldn't have seen the misinformation machine that Facebook was going to become, with like all these bad actors and all you know, sort of the ways that you could abuse the systems, Like we didn't know that that was going to be a thing until like we started to see the actual abuse. How do you protect against the things where you've got to take in like the worst of humanity? Like, are you doing that on an active basis?
Right?
Because like the thing with a tool like this is that the best parts of humanity will find like amazing things to do with it. But there is an equal opposite actor there, right, who will do the worst things with it. So tell me, like how that you build a product like this and don't let it become destructive.
Yeah.
So my life philosophy is that creators imbuing their values and the things they create, whether they know it or not, and that those things have a way of spreading those values even when they're no longer around. That does actually put a lot of blame on people like Zuckerberg. It implies that he made Facebook with the wrong values. I don't know, Mark, But an interesting example that I like to think about is the defaults of something like Facebook versus MySpace. Like obviously he was aware of MySpace. We know that, right, definitely with the main compeditor.
Yeah.
And when I remember going out to MySpace for the first time and my page was blank and it' said I had one friend, I was like, who's my friend? Oh my god, it's Tom, Right, Who's Tom? You know, it's this nice guy he's the guy maker of MySpace. This is cool, Like Tom's my friend. He must care about me. I bet I could make other friends that I don't know, like my face's places where I can make friends, right, and Tom cares. And when you sign on to Facebook, you have no friends and markus certainly not your friend.
I never consider this. But he's not your first friend on Facebook. That's definitely like what the fuck does that mean? What the fuck does that mean?
Not only is he not your first friend, but you have no friends when you join right right right? When you join Facebook, you are this friendless non person then, and you have to try to grab out to anybody who you already know, like please, somebody who already knows me be my friend on Facebook? Right?
Interesting? Huh?
And like these there's like these really deep details that are made by real people who have values, Like he had to think about this, obviously he thought about Like he's not dumb, like he must have thought about it.
I mean maybe he wanted to be your first friend, but they were like actually, like MySpace, Tom could sue us for like IP stuff like infringement if we do the same thing that he did.
You know, I think we know he wasn't that cautious about being suited because it happened, right, Uh, that's true.
I mean.
There's a lot of interesting things like that. I think that actually maybe everything is that way. The goal is like not to not make things, but to make things with like really good values and and to have people with good values making things, and like that making things is not equivalent between any people.
I agree with you, but like what is the expression, like the road to Hell's paved with good intentions? Whatever? I mean, I agree that that you can avoid some of these mistakes like if you have a different set of like goals or values, But like do you already do things with mid Journey where you're trying to sort of protect against like misuse, right, Like yeah, obviously like hate speech or images of violence. I mean I definitely like tried some stuff that I didn't think was like going to produce a violent result, and it was like, we don't do like this kind of image or whatever.
Yeah.
I actually have a question about porn, which is a big one. Go ahead, Yeah, I mean my guess is if you wanted Mid Journey to create like incredible original like porn scenes, because there's a lot of pornography on the Internet, right, would you say there's quite a bit of it, and it's all a visual medium basically. Yeah, I mean there's obviously some erotica out there. There's somewhere You've got the X rated Mid Journey instance running right where I can create like full on porn scenes, right, don't lie to me. I know the truth. Somebody there is doing it.
Yeah.
You know, when I first thought about this problem, I was like, who wants an AI generated booty?
Who doesn't?
And then like, honestly, as the albhims get over time, like I see some booty, then I'm like, it's a pretty nice booty, Like it's pretty good, pretty good. Yeah, Like it obviously can do really good like just how you can make beautiful anything else.
I mean that's a huge deal though, Like I can't even do like a Renaissance painting of nudes like tasteful artistic news with mid Journey correct, No, right, Like is there a tier where I can do nudes? This is really just I'm asking for myself. But like, you know, no, you're not gonna let anybody ever do a nude.
I you know, I think it's about like, what is the thing that like helps the world, like what and so like. For example, there are two things we have tried. I can give you two stories. Okay, Well, one is when their system wasn't filtering well enough, you'd have people trying to basically create like their fantasy person basically and they're like becomes super fixated on like this redhead whatever, like it becomes this very specific thing over time. Right, I don't know if that feels healthy. It's certainly a market.
Right, I mean, by the way that phrase, I don't know if it's healthy, but it's certainly a market. Is like ninety percent of the things that are available online, like literally social media is like I don't know if it's healthy, but there's certainly to.
Someone's going to do it, and I think it's not going to be healthy right now. There are other things that we tried. So for example, we did this thing where we created this chat room and we called it not safe, don't judge, and we threw like a hundred people into it, and we turned off all the filters, oh my god, just to see what would happen, Oh my god, And it was really interesting. We put them all in and we go, there's no filters everybody. You can do whatever you want, but everyone else is going to see what you see. There's got to be some people who would be shameless in that scenario. It was very quiet at first, and then someone goes boobs and then there's some booth pictures and someone goes like ass and that it was a good ass picture, and everyone's like kind of startled it for us, like they didn't know what to do, right, and then somebody goes, uh, fifty percent orgy in a Walmart and it just like these piles of good bodies in a Walmart sounds very disturbing. And then all of a sudden, everyone else goes, uh, it was fifty percent orgy in space, alien orgies, and then all of a sudden everyone starts losing their minds and it gets really strange. Eventually it went to like Bill Cosby eating out Hitler, Like it got pretty intense.
Oh my god, I mean that's a very, that's a very, that's a full cancel on that image. I would say, yeah, everything, But.
What was happening was like it became so absurd, Yeah, that everyone just started to kind of like let go of all of the bullshit that they knew that like that they would normally be outraged of. And when somebody finally did Bill Cosby eating out Hitler, like that was like an hour in okay.
Yeah, and is that when you shut it down? Was that this was that when you closed that?
I shut it down shortly after.
Yeah, But that's like such a small sample and like it went immediately to a place that would defend like probably ninety nine percent of the normal users of the Internet.
But I think what's interesting to hop about psychological experience of all people had in this room as it went from like boobs right to like, you know, they kind of escalated to like fifty percent orgy in Walmart.
Isn't that what always happens though, like you're testing the limits.
No, but like no, but what happened was it's like at some point they kind of like let go right during this process and they were like it doesn't matter anymore. Yeah, Bill cosbyah Hitler, that's really funny. Or someone else did like Michael Jackson's asshole and it did like a buttthole where the hole was Michael Jackson's face.
It was funny, it was weird, you know, those people.
Thought it was funny, but like a very large audience would not think that was funny. So the thing so like, I mean, it's not funny, like at it at a kind of basic level, like you know, the Cosby stuff is really fucked up and Hitler is Hitler. So like at a really kind of basic level. If you're like in good taste, that's very very not in good But there was no taste anymore. It was like everyone just like lost it.
They're like, look, nothing matters, like it's all bullshit, like it doesn't really like everyone kind of let go. It felt very ethartic. At first, they were really shy, and by the end they had all let go. Yeah, it was kind of a beautiful process. I don't know though, but like it went to a place that was pretty offensive, right. I mean, I'm glad that you don't allow that particular type of use in the broad I think it was really interesting and I would say everybody who was involved in the Spear and stelt it with like cathartic, right, and a positive like spiritual experience, right, because they realized how pent up they were in stupid ways and like maybe the last thing was bad, like we could.
Say that was bad, but there was something.
No the last thing was bad. The last one was bad. I don't I don't want to be like a you know, like the policing culture or whatever. But I mean you could do it, but no. But the reality is like, actually like it. I think that raises an interesting sort of scenario, And it's like, what do people do when given this kind of unbridled power to create whatever's in their mind? Like I like to think people come up with like really cool stuff that's like awesome, but definitely for sure, there's a segment of the audience. And this is actually gets back to what I was asking, which is like, so you ran an experiment with a roomful of people. They were just like users, like test like beta users or something.
There was a bunch of users.
We did it for one hour, right, and I said, if anybody leaked an image, I would ban them for life.
Right. So that's your little kind of window into it. You're like, Okay, this could get pretty crazy. Oh yeah, obviously, the way you've built the system is that you cannot do those things. I guess. Like the question is like do you have to be constantly vigilant about like the ways that the thing might be abused? Like how do you counter like abuse you haven't even thought of yet.
We have like forty moderators who kind of watch things and then they just they have they have a little slash band commands, so they say slash band titties and also, no key is the word titties anymore?
Right? Are you actively like yesterday, was there something that mid Journey produced that was like a surprise to the moderators.
I know that there are words that were banned today.
Like what what was banned? I'm super curious, Like, but today you're way far into it. There's like, how many people have used mid Journey? Do you know the numbers? Million, millions of people. Yeah, so millions of people have been in there. But you're still today as of October fifth or whatever. Yeah, you've banned words. I'd love to know what the last band word was.
Moderators came back recently and they're like, David, we want to unband the following words blood, bloody, sexy, kill, killing, cutting, disturbing, and gut.
Wow. What an image.
They're like, what do you think, David? We could probably unban those things. And I was like, Okay, let me think about this. Uh, child with guts, Bible across the ground, disturbing huge pools of blood and like, ooh, yeah, we probably don't want that right where I was like a little girl cutting themselves, Like oh yeah, that seems bad right.
Well key, But so here's my question for you, And I think you've got like kind of a crazy responsibility. And I'm not saying this to be a jerk at all, but like you're just like a guy who's interested in creating this product and create this kind of beautiful and imaginative and exciting images and beyond. But you're not like a linguist. I don't know you're all of your background, but I mean, like you're not like an ethicist. Do you employ an ethicist at the company? Do you employ like linguistic experts? How diverse is the team? I think these are like things that people are going to want to know, which is like you mentioned the Bill Cosby Hitler thing, and I can think of like a bunch of people who are not like a white Jewish guy. And I say this as a white Jewish guy who would be much more offended about some of that stuff, or people with different experience.
That as an example of a pretty outrageous thing that was my outrageous and I get that, and no, no, I understand it.
Like you were in this experiment, somebody took you to this crazy place and then you're like, all right, we got to shut it down. This is sort of what I was trying to get to is like, how do you make a company that has all the lofty and interesting and exciting ideals I think you have, but also protect against building a product that ultimately ends up repeating the mistakes of the facebooks or the twitters of the world. And the question does come down to, like when you're having those conversations, who's in the room, who's having that conversation with you? Like, what are you going to do? This is my being putting my hardcore journalist hat on, Like what are you going to do to make sure that you have conversations with a big enough set of people and with a smart enough set of people who are experts in these fields, like in the fields of like ethics and linguistics and like you know, history, and that it's a diverse group, like to actually make a product that serves everybody and not just one that feels like cool to like a couple of you know, Jewish guys, like us, but may not work for a million other people in the world.
Yeah, I mean there's a lot of questions there. I'm okay not serving everybody. Like if this is I Maginty's a two million person thing, is ever bigger than that, I'm.
Happy with that. But you want to make it inclusive, I would assume, Yeah, I want it to be inclusive.
But also if it's only two million people, I'd be okay with that, Like I don't, like, I'm not I don't have this like this desire.
You want two million of the same people, though you don't want two million to the same people, two million white Jewish guys.
If it makes two million white Jewish guys really happy and improves their lives in a significant way, like they've made the world better. Now, obviously I'd like to make it diverse, like and we try really hard there. But like I mean, at the end of the day, it's it's more important that it's good for the people who interact with it then that it has as many people as possible.
And that's the first trade off.
That's the first that's a huge trade off, because most people decided to not make that trade off.
No, I agree with you that, like, if you're thinking of like the infinite audience, obviously you don't want to be like every person should be in this thing or using this thing or whatever. But like, I guess it's such a sensitive space where like you've built a tool that can create something out of nothing, Like you build a tool that can make a dream look like real basically, And so yeah, you know, how do you do it the right way? But I feel like here's a chance to bring a bunch of people into the conversation that were never there at Google on day one. When I think about any new technology like this, I always think now, and perhaps because I've been so abused by the technology companies that have existed before us, you know what could go wrong? Right? And how do you prevent that?
Yeah, there are a lot of things we do, so like I do office hours every week for four hours where I just talk to as many people as I can. Sometimes I'll do a theme thing like I brought up like twelve women once and I said, like, let's have a women panel, and I want to ask everybody how do you feel about bikini photos? Like should I ban bikini?
And that's one way of getting the women's side of things.
Because every single day I heard some asshole dude who's like, hits are natural. I like, bikini photos, have as many as you can, and then like women who are uncomfortable, And I was like, you know what, I just want to hear a bunch of women talk about this issue of how do you feel aout the bikini photos and like I will do whatever you say.
I was like, should I ban bikinis? That was like the simplest question.
Did you ban bikinis?
They decided group like we do not want you to band bikinis, like ninety five percent. It was like pretty unanimous, but we want you to hide them so that none of us ever have to see some dude making a bikini.
And so that's what we did.
It's a good middle ground. To me, this is so weird because like the reality is like the naked human body is that I'm not like, on its face offensive to me in any way, Like it's like very normal.
I agree, Yeah, And.
It's like funny to think that you've got a buffer against like people abusing the system who are making weird like you know, sexual bikini photos or whatever.
Yeah, I mean what the women basically said on the on the whole is that they're like they're basically even we like a little cleavage, but like.
What an average guy thinks it's sexy.
It's really easy for most of us to find creepy and unwelcoming, and so basically we don't have we don't feel like we should have to see that like against our will.
That's so true both in AI and in life.
There's a lot of these sort of nuanced things like technically, it probably should be able to do a tasteful nude, but it shouldn't be able to do like a hyper sexualized nude. Like technically like that seems right, you know, but it's it's hard. That's a really hard boundary, you know. I mean it's a question of art, right, Like, yeah, what's porn? It's like, well, you know when you see it, and it's like, but there's different levels of that, right, have we even trying to teach the system? Actually lately some of these nuances. We have certain users who go in and they rate images randomly. Right, We find that on the whole, people very rarely say anything is offensive, like very rare, so when they say it, it's interesting, right, And then we and we aggreate all those together, and then we teach the AI. We're like, hey, regardless of whether or not something is offensive, this is how people are responding to your images.
Interesting.
And then what it does it actually changed its behavior?
Do you worry you're you're creating a kind of prudish AI? Like, do you worry that, like you're actually making a sexually repressed AI that like is going to be weird about sex and human bodies.
I think the question is, like when we build these technologies, like what battles do you want to fight?
And where do we want to push the world forward?
Right?
And like me, I want the world to be more imaginative, like and I want to push the boundaries of like aesthetics in creation. I think that's really interesting and it is really worthwhile. But I can be a little picky. I'm not as interested in doing that or violence sexuality. Right, there is an argument we have to push the batteries of sexuality. Let's make the world way more sexual. Someone else can do that. I just don't feel.
Spiritually compelled for that.
Yeah, But I think that, like there's this broader thing, which is like letting people reflect. The average person comes in here and they say something like Maltese dog in heaven and I reach out. I'm like, hey, why'd you do that? That's interesting and they go because my dog just died. And I'm like, oh shit, are you okay? And they're like yeah, this is making me feel better. Where there was like another woman and she goes like she was putting in these weird lyrics and I'm like, what are you doing? Like these don't show up on Google and she goes, When I was very young, my older brother died and all he left me was this, like this cassette tape.
Of these thongs.
And I'm literally just putting lyrics in and I'm feeling closest person never got to be part of my life.
Wow, it's not always death.
There was one person who was like Temple of Donuts, Like why are you doing Temple of Donuts? And like, well, I'm an atheist, but I don't really understand worship or religion, but I do understand like donuts and sweets. It's like combining all the things I don't understand one of the things I do understand, and I'm like trying to understand, like what is worship?
The Hong Kong girl. So she said, I'm a woman.
I'm in Hong Kong, and the one thing your parents talk, I never want you to be as an artist. And so I'm a banker and I'm a good banker. But now as I'm starting to get to use Mid Journey, I'm starting to get to feel like I'm getting to be the person I never got to be, and I'm having to think about that.
And so like these are like the good stories.
They're like, no, those are great stories.
Somebody else is.
Just like huge chits covered in blood and it's like I don't care about that person. That's not a real human story, and like maybe there's something going on there, but it's not interesting. Like there's so many interesting things going on, and I want to create a space for that, and I'm doing that. There's a path that we see over and over again with people in Mid Journey almost call like the heroes in Mid Journey. And what happens is they come in and they realize they can make pictures of something they like. For me, with cats and cyberpunk, right, I'm like, okay, I make sappunk cats.
I'm like, okay, I'll make tarpunk shitte.
Pre Charpuk Ninjas and make Starapunks and I'm making chappok everything. And then all of a sudden, like you combine all the things you like and then you just burn out and you're like, oh my god, I never liked sypunk.
I never want to see starpunk again. Starpunk isn't me.
And then and then it's like month one, Month one, and then month two is You're like, but then who am I? And then you start looking at everybody else's pictures. You're like art deco, am I art deco? Or like vapor ways, am I vapor waves? And then you start like looking at everything and you're and you're kind of saying like you know, and you're trying to do this path of like who am I?
What is my real aesthetic? And then you learn a lot.
People learn like all this hard history and all these movements, and they start putting things together into like the sense of who they are. And then like month three is like you have this like aesthetic universe and you're starting to like apply it to everything. You're like, it's like, you know, it's a little bit of this, a little bit of that. It's all these things together and you're like creating all this stuff, and it's like you're like having to like your people are paying.
Off pathetic debts.
They're like exploring the nature of their identities and then they're like expressing it. They're like it's like they're working on all this shit, and like it's really really healthy and it's literally just regular people and almost nobody shares anything. It's crazy, like almost no pictures ever shared and almost no pictures ever sold, right, And it's just like it's mostly just regular people having this like really healthy experience.
So to be clear, basically you see this as a form of therapy. Is that correct?
At least thirty percent of all the use is literally art therapy.
Right, Wow, mental health through ai I was.
Entirely unexpected, but it's really important. It's clearly this tool for reflection. And then people are starting to meet each other like and they're starting to like form these groups and they're like pushing these aesthetic boundaries and discovering new things and like that's really beautiful and it's obviously part of an honest and positive future, right, and like that's what I care about.
Okay, really quickly, and then we got to wrap up, But do you think that like there's a future state where it's like mid Journey is its own Instagram.
There's gonna be like that, but it's crazier. I think the future is more of like, well, it's more like liquid imagination swirling around the room and like forming mountains and little trees and animals and little ruins. You're trying to figure out how to get people's surfboards or boats like surf like oceans of liquid imagination, like discover entirely new lands. But it's like very different thing, and it's like it forms like a new substance that you kind of can create the world with and manifest through and like reflect through, and like that's what it's about.
It's like creating a new substance.
It's really not about like making an Instagram or making poor or huge tips. It's obvious that all that stuff will happen, but that it doesn't matter, Like it's not the real thing. It's like there was a civilization before engines and after engines, and now the fun thing is moving to a civilization that has these engines of imagination, and how does that transform things?
Like how did engines stright?
I think we have highways, we have boats, we have like huge international trade, Like there's.
Like a lot of stuff.
Yeah, that's all stark. A lot of people in technology feel like we have no pasts. A lot of regular people literally feel like we have no future, right, But like I feel like like we're really mid journey in this, Like we have this rich and beautiful past behind us and this like wondrous and unimaginable future ahead of us, right, And like the whole goal of making anything is to figure out what we can be and what that can be in like a positive and explorative and wonderful, humane way.
And like I don't know. That's what I'm trying to do and hopefully it shows up a little bit in the stuff.
But like I agree, I'm so on board with that sentiment. Like I one hundred percent agree, Like we don't know yet what all of this is going to be be. It's like we have to figure that out. And that's why like people are like we're done, and it's like, no, we just really got started, like with technology, I think, and what it can do. I agree, Like you, you are echoing a sentiment that I have definitely spoken on on more than one occasion. David, this is first off, super fucking interesting shit that you're building. Extremely fascinating conversation. We should do like a check in like a year from now or something to see all of the new mid journey things that have been created.
So it's gonna get really scary. Even the next six months, six months is going to be really intense.
Like six months is the farthest I can see twelve months actually, I actually don't.
Okay, we'll do a six month check in. We'll see if bid journey is that it's like three quarter journey.
It's gonna be.
Yeah, it's gonna be.
It's it's gonna be moving really fast.
It's kinda seemed frightening to a lot of people, but it's like it's it's like an honest shot at the future.
You know.
I'm ready, David, Thank you so much. Well, that is our show for this week. We'll be back next week with more what Future, And as always, I wish you and your family the very best.
