Generative Art Assets for Games: Emmanuel de Maistre
Emmanuel de Maistre ("Emm") is leading Scenario,a platform for creating game art for using generative AI. One of their unique approaches is to focus on helping developers build their own fine-tuned models based on their own art style--helping to establish aesthetic consistency.
This is the second in my series of conversations with Generative AI pioneers.
During this conversation we spoke about the future of the game industry (across many forms of generative AI beyond art), Emm's background as an entrepreneur, and even got into some generative applications within the sciences & biology.
Show Notes
Some of the articles and companies referenced in the article:
Jon’s Computational Creativity article
Jon’s article, The Direct From Imagination Era Has Begun
Emm’s company is Scenario
Jon’s company is Beamable
Transcript
This is an automated transcript generated by Reverside:
Jon:
I am with Emm, the CEO of Scenario, an AI company that is helping anybody create game assets using AI. So Emm, first welcome, but why did you create Scenario and what are you most excited about in this industry?
Emm:
Hey Jon, thanks. So my name is Emmanuel de Maestre. People call me Emm. And we created Scenario two years ago actually. I got absolutely fascinated when Apple released a LiDAR sensor on the iPhone that was about two years ago. And I realized, especially thanks to my previous company, I realized people could be able to scan and create 3D very easily with their phone by scanning, by pointing the phone at something, the 3D model of that something in the phone. And that was the first product we made for about a year. And then Jenny and I started to eat up about like six, nine months ago. And I realized that the new way to create stuff, to create assets, to create visuals and 3D models would be by prompting, which might be complementary to scanning, but so much easier since you can generate assets anywhere by just describing what you have in mind. how we started the company from riding the wave of 3D capture and then transitioning to GenEI.
Jon:
So how is the whole market for game development going to change?
Emm:
It is going to change because it's going to get so much easier to create all the content you need to create a game. And that's not just images. I mean, people see today Mijoni, Dali, Stable Refusion, and others and us. That's definitely the very attractive parts, because these nice, fancy images look very cool. But all other gaming content can be AI generated. animations, probably 3D models, certainly in a few months. Entire, maybe ideas, scenario of games, videos can be AI generated. Look, it's like the whole different categories of content will be AI generated. So we should expect way more creativity and faster processes to create a game, from idea to shipping, launch.
Jon:
One of the really interesting aspects of scenario is that I can go in there and train my own model as well. So this isn't just like mid-journey where I go in and I prompt it from this massive database. There's this whole element of like I could give my own art. What's the thinking behind that? How will that change the production process? So I'm going to go ahead and start with the question.
Emm:
I mean, that's what I heard from customers or users, games to use specifically. They are looking, or the industry as a whole, is looking for the highest style consistency, not just the random exploration that you can get, sometimes great exploration, but still a bit random by prompting. So style consistency is key. And training your own model nails the visual feature that you need when you have a specific style, specific art, specific graphic direction. And beyond style consistency, studios are looking for a lot of data security constraints or very clear frameworks. Data privacy is also very important. Some lot of studios for a good reason are somewhat worried or wonder, you know, what are the legal implication of the databases that have been used and so on. So style consistency, data security, data privacy.
Jon:
So just to kind of drill into that for a moment, their concern is maybe that they're utilizing someone's intellectual property off of someone else's model. Is that the core issue? And what's the privacy issue on their side that they're concerned? And so I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point. I think that's a good point.
Emm:
Yeah, so the first question, yes, they are concerned that some somewhat like using databases that are too large might be infringing some copyright laws. Data privacy is when all the content they will generate. I mean, they want the content they generate to be extremely private and to stay within their own company servers accounts. They do not wish for that content to be shared, reused, retrained. And today, on some platforms, everything is pretty exposed. Like if you look at Meet Johnny, which is amazing, largest Discord server in the world, everything you prompt is available for everyone. And the data is being reused by Meet Johnny, which makes the software even better.
Jon:
Right, but if I'm a big game publisher, maybe not even a big game publisher, I'm gonna wanna train my model, get it really perfect to my style, and I don't necessarily want other people being able to access that and build things off of it. So that's sort of the privacy aspect, is keeping it inside my company.
Emm:
Yep. If I may elaborate on that, people will want people, game studios, some game studios or game companies will want to bring content creations in the end of players. That's like UGC, player created contents, it's big. It's not for everyone, but everyone wants a piece of it. Using AI and using your very own models, very own generators is a way to control the style and bring generation to your players by ensuring the style stays consistent and is respected. And that's what a lot of studios are interested as well, not beyond what I mentioned about privacy and security.
Jon:
So I'm imagining the next generation of sandbox games, like we have Minecraft and things like that now, but the next generation could actually be, not only are you shaping the environment, the player, the creator is actually coming in and prompting things into existence that actually match the environment that's been set up. Is that sort of the need that you're hearing? Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah.
Emm:
So it could match the environment, which would be about visual features matching the style, but it could match the story.
Like you could train a GPT 3.3.5.4 model based on your story for players to create new content, new texts, or new dialogues that match the original story. If you are in a medieval game, do not expect cyberpunk, sci-fi dialogues, right? So like hyper-personalization, your AI I think will be key for game development.
Jon:
So the players now are really co-creating a world with the developer in that particular vision, which I think is fascinating. What is this going to change about smaller game developers? Like will smaller game developers be able to compete with like a AAA game studio?
Emm:
That's a stretch, but eventually, yes, why not? I mean, why not? I'm seeing teams of like two, three, four people developing some very interesting games and ideas using JNAI. And because they can get their hands on art, which they weren't able to before, they can go further away. Like one example is I have a team of two people games, a rate-wrap-through game, and they are ready to bring character generations in the hands of thousands of users that should do at least 100k to 200k characters per month.
Try doing it with artists like traditional processes. Try creating 200k characters by asking the players to go into Blender or Photoshop or whatever software that's impossible.
But with the AI, creating one image take no less than five to 10 seconds.
Emm:
And it should cost a few cents, maybe a dollar, maybe less. So that's why GNI brings much more power in the end of Indies.
Jon:
Well, there's something really important at the core of what you're talking about though, which is, yes, it's going to be a lot less expensive to create massive amounts of assets, but it's not just about having more assets and more quantity. The gameplay itself sounds like it really could change a lot with these kind of capabilities. When you can start actually having these assets generated in response to how you're playing, That's a new kind of game. That's these these creative, co-created games with the player.
Emm:
and it starts with the assets today, but tomorrow the asset will be tight, associated with text, with a story. The asset might be animated. The animation might come from the creativity of the player, so it can go pretty far, really. And the studio, the game that can leverage that player creativity wins, in my opinion.
Jon:
And every day I post about this stuff, you must see way more than I do, but I'm just curious how you think about it. But whenever I post stuff, there's always the inevitable comment, which is like, oh, well, I've been training as a 3D artist for years. Am I out of a job now? What do you want to say about that?
Emm:
Yeah. I mean, I see the comments as well, and I understand the concerns, for sure, even though I've not been trained as a 3D artist myself. I can definitely see why the other concerns are here. On the other side, we have more than 20% of the scenario users define themselves as artists, 2D artists, 3D artists, game artists, whatever. Like we specifically asked, we did some surveys, a few thousand people answered, 20% are artists. And I'm getting feedback from this category this is the best tool that they discovered so far recently to produce, to create, to really like develop their skills.
So it's, you have as an artist, you have to go beyond the initial fear of the AI taking your job and consider the AI as a tool to help you make an even better job. the same quality and do better and do more. Or do different. I mean, it's not just about more and faster and cheaper. It's also like different. So, yeah.
Jon:
It seems like if you're really smart about how you approach these tools that you could really multiply your productivity immensely like you could now be like a creative director and and figure out the style for your game and that could be based on skills that you've invested in right it could be your own illustration style and now you instead of hiring 10 more people which are hard to find to recreate your style the AI gets trained to it and it is your style it is your intellectual property that you can use. that you're bringing to life at a scale beyond what you would ordinarily be able to do.
Emm:
Yeah, you might even develop new styles that you would maybe you would not have the time to develop yourself, because it takes time, it takes energy to create new directions. And so you can use AI to explore faster and decide what that should be your next style for the next game. And you know what, like, artists are the best position for these technologies, at least for, let's say, visual image or 3D generations. Why? Because they have the culture, they have the language, terms. They know the words that are necessary to make it work better than others. They know how to differentiate different shades and lighting and composition of an image and so on. I'm not saying no one else can do it like people
Emm:
can learn but artists have the skillset necessary to make it work better and faster than others so they should leverage it before somebody else tries to do it for them.
Jon:
I'm reminded of this section of Richard Feynman's book. It's called, Surely You're Joking, Mr. Feynman. So he's a very famous physicist, Nobel laureate. But there was a time in his life he went and trained to be an artist. He thought he couldn't accomplish it at all. There's a couple of interesting pieces of it. One was actually he proved to himself that if you are dedicated enough to something, like no one lacks the capacity for art. But what he also learned in the course of this was a really deep appreciation for art, and he felt he could now go out into the world and really understand a lot more about what the artist was trying to accomplish. So the reason why you just reminded me of that is artists who really understand composition, who understand intention, who understand what details are really bringing the narrative of a game experience seems to me that they're going to be super powered now because they're going to really be able to sculpt this art into existence in a way that they're multiplying themselves, they're getting more content, but it really matches the intentionality and the artistic spirit that they brought to it.
Emm:
there should be power. And there should be always a significant human involvement, whether it's for guiding the AI, for polishing the output of the AI, for combining different AIs together. We should not expect games to be AI-genoted by just pushing a button at all. But it's a new skill set being developed, and I think artists will develop it faster than others.
Jon:
So right now, when I use scenario, it's 2D art. And earlier, you were talking about all the other asset types we were talking about, music and levels and 3D graphics. Well, I think I heard you say 3D graphics maybe. What's hard about 3D?
Emm:
It's very hard. I know it because I've been in 3D for 10 years. First, scanning with drones, large areas. Then scanning with phones, which is a different sensor or different device. 3D is messy when you have to combine the mesh itself. Sometimes the rigging of the mesh, how it moves. And then all the texture and materials normal sense and so on. It's pretty hard like versus like a PNG or a JPEG. That's only like two-dimensional. But if you look at what so that's why I'm saying it's hard much harder than to the visuals but it's coming.
But everybody except everybody, like most people that speak with expect AI-generated 3D to be real in games within the next few months, like six, like three to six months. Of course it will depend on which game. Are you talking about like a casual 3D game on mobile, triple a game, like a Minecraft sandbox, you know, Voxel sort of game, there will be a step-by-step process. We should not expect, you know, Assassin's Creed to be AI generated yet or,
I'm not sure, not sure. But like, look at Roblox is moving pretty fast in the space. Sandbox just announced the partnership as well and initiative recently. You can generate the texture without necessarily generating the mesh as well. So there would be ways to get into 3D AI generated 3D.
Jon:
Yeah, I posted about a year ago, I predicted that someday someone will be able to just sit down and make an Elden Ring. You were a little more skeptical. See it on your face. You're a little more skeptical. You've touched a lot of this stuff. I feel like it's accelerating so fast. Maybe it'll be 10 years, which was my estimate. Maybe it'll be a few years. But it feels like eventually we're going to get there. But there are so many steps in the process. Like if you think about going from concept art to the to then the textures and the UV unwrap and UV unwrap. I've yet to meet an artist who loves that part of the process. Like if that could just go away, so many people would be happy. And then rigging it, that's also pretty annoying to do. And then animating it and getting that right. And then working into the physics, like so, so, so many steps. It feels like there's just, there's so many problems to solve and there are very different problems as well.
Emm:
Yeah, but think of like two years ago, nobody was expecting 2D, like AI-Jet 2D to be at where it is now. With Mid Journey, V5, with Stability Fusion, Excel coming with Dali 2 being so advanced and more coming, nobody would have expected that two years ago. People knew it was coming, but at that level of quality, in the next, like within such a short timeframe, no. So the same might happen in 3D. Like other media might be easier to generate video. It's coming, look at corridor, how they made an anime video just by shooting on the iPhone. Then audio is making some really nice progress as well. Text obviously with GPT-4 and others coming. So yeah, everybody's dreaming of 3D as the Holy Grail. But other part of, you know, Gen AI will move fast in the meantime anyway.
Jon:
Now, we've been talking about games, and games is a great industry. It's 300 billion plus and growing. So if all we ever did was automate all this aspect of games and add new gameplay functions, that would probably be great. But it just strikes me as we've been talking about this. There are so many other uses for this whole 3D graphics pipeline, everything from movies to, but all the real time stuff that comes from games, like simulations, all the quote unquote people have been talking about won't all of that be built off the same text?
Emm:
Yeah, well, it's different like metaverse have been in that you've been in that metaverse space for longer than I was. By the way, I was using your own, you know, medium article and post to pitch VCs when I was starting scenario. metaverse is like it is it is going to be a reality that everybody recognize the struggle for the metaverse is like the quality and the quantity of experiences that can be built in the metaverse. And it is very hard to build a lot of them. So if these experiences can be anyone using JNEI, we might have a key to really make the metaverse a reality. Beyond infrastructure, beyond connectivity, and you spoke about the layers of the metaverse, right? You can elaborate on that more than I can, but creating the content and the experience can be now easier thanks to JNEI, or will be easier thanks to JNEI. I'm going to start with the first question.
Jon:
Yeah, I mean, I think the whole thing that we call the metaverse is really from a technology perspective, it's all just based on game technology and applying game technology to a lot of other kinds of experiences. So you could take like music, for example, like music as an example, because in Roblox and Fortnite, we've already seen like music concerts where there's been tens of millions of people who who want to experience that. And it's in a game engine. But it's not. per se, you're there to experience things. So we're going to see more and more of that kind of thing, and it'll become more and more accessible to creators to enter into those environments. So I think that's exciting as well, because so many of these pieces, you're just going to want to generate them. Like, let's put this in reach of everybody.
Emm:
You know what? So on Friday, we announced that we would work with Sandbox to create new characters, words on Sandbox. Well, Friday wasn't the best time to announce a deal, right? Like it was the day SVB crashed, or shut down, at least. So I was pretty worried that the news would go unnoticed, which would have been understandable. And we Sandbox announced they would try, and love for users to get on board and try this new approach of creating assets in the sandbox with scenario. And within four days, we've got 30,000 people registering for the beta program. In four days, like 30K people saying, yep, I wanna be among the people creating the new assets for sandbox. I wanna be among them. How enormous is that? So now apply that to Minecraft. Apply that to other, maybe rec room, maybe other player generated content environments, that can be big.
Jon:
This is one of the pieces that sometimes gets lost in conversations around these tools like ChatGPT and whatnot and mid-journey, which is that it's actually just really fun to go in and do this stuff. People who have tried it know it's fun, right? You go into ChatGPT, I made so many different things in ChatGPT. It's just fun to see what it's going to do. Sometimes it gives me other ideas, but it's the experience itself and seeing what it reacts with. The same thing in mid-journey, like what is it gonna do next? to whole virtual worlds and creativity, like there's the aspect of playing in the universe, just like being a gamer, fighting enemies and all that kind of stuff that we like from games, but just the act of creation and shaping things, that in itself is there's such a like a pure joy to that, that I would love everybody to experience.
Emm:
And players are, I think, creative people by nature and game creators even more. So that's why maybe this category of people should enjoy JNAI way more than anybody else. And there is a symptom that I've seen. I'm not sure it's been described by doctors. When you start on a mid-journey, typical mid-journey account, it could be Delhi, Stability Fusion, whatever, when you start, it's so mind-blowing that a lot of people get stuck for a week and basically don't sleep. Like sleep a few hours per night. just because they're so mind blown about like the, you know, discovering this new process of putting ideas into images in a few seconds.
So I've seen persistent testimonials of people on Twitter, you know, explaining how little they slept the first week they discovered mid-journey and the likes.
Jon:
Sounds like the week I discovered World of Warcraft or something like that. Yeah, it's the ability to just create things. I think of generative AI, especially with the LLMs like chat GPT, it's like those are virtual worlds. Like they're not hugely detailed virtual worlds at this point in terms of the persistence of them and the ability to navigate and everything stays the same. But whether it's this guy's demo where he made like the internet and Linux virtual machines in chat GPT. And then you could actually log into the shell and like log into websites and see what, it's like that was a virtual world or what I did. I made like text adventures in chat GPT. Again, it's a virtual world engine.
Emm:
Have you seen people creating Pong with GPT-4?
Jon:
I did.
Emm:
In just like, how long did it take? 60 seconds,
Jon:
Yes, I did see that. Yeah, I'll post that link in the show notes because it's pretty mind blowing demo, but yeah, a guy who I think he worked for Brex actually, he went and was messing around with it and he told it in less than 60 seconds how to recreate the game of Pong. So you could imagine how this is going to continue over the coming years.
Emm:
you know, creates fortnights with GPT4. Let's do it. No, it's coming fast. It's coming definitely very fast.
Jon:
I'm super curious about your background because you've done a whole bunch of what at the surface to me look very different. You've been in biotechnology and then you were doing drones. And I heard what you said earlier like LiDAR and drones and 3D scanning kind of brought you to this. Can you look, can we talk about your background for a little bit? What has led you through these diverse fields? Is there a common theme behind all this?
Emm:
Yeah, that's right. The common thing I think I'm fascinated about like technology being used and applied and used by humans and applied to humans for the greater good. Biotech, I've been trained as a molecular biologist.
I love the tech itself. It's insane. By the way, I wish I had GNAI back then, you know, when I was sequencing DNA and like trying to reproduce like the structure of a protein and so on, like it's so much better now. And I went into biotech because I felt all these great advances in biotechnology where molecular biology were underused. That was like the early 2000s, right? In the meantime, I was pretty heavy gamers back then, doing all the world of Warcraft, lots of Starcraft, lots of real time strategy games, like all of them spending nights playing. One of my biotech companies, one of the biotech companies working for got acquired and I decided I would take a break and become a pilot. Just for the dream of flying smaller planes. The same, yeah, bear with me. The day I got the license, Europe allowed private pilots to fly commercial drones. Back in 2012, which was ahead of the US back then, the US only unloaded that in 2016. And so, yeah, in a new toy, new technology, I had to put my hands on it. So I started flying drones. which was pretty, pretty unusual.
And some friends heard about it and they asked me to take pictures for their own construction sites or infrastructure and so on. And I realized the regulation was opening a new market. And so we ended up doing a company doing drone data analysis using AI to identify specific feature within the maps and 3D models being constructed. And we ended up setting the company in 2016 to a US data. startup in the Bay Area, and I moved to the Bay Area in 2017. And looked like a lot of the place. That's a place where a lot of innovation happened. So I stayed, and I decided I would do another company here, even bigger,
I wish. That's the story. But I'm fascinated by tech being applied in real world use cases. And I feel Jenny and I is just this. The tech has been here for five years. The tech, the concept, aren't quite new, especially transformers, like for texts, maybe it's newer for images and 3D and so on. But the tech has been here for a while. Now, today, this year, 2022, 2023, is the year where this tech gets mainstream in the hands of billions of people.
Jon:
So we share some interesting interests. So over a decade ago, I was a teaching fellow for George Church, who was teaching a computational genomics course. So for those watching this, and you didn't know we were gonna talk about biology, George Church is a very famous biologist. He has done a lot of things. Right now he's trying to bring mammoths back to life, basically, but he's done all these things around how you apply computer technology the genome and what I had been always really interested in was, well, could we do drug discovery almost automatically if you understand biology?
Now, if they had had AlphaFold back then, which so AlphaFold has basically decoded the proteome and it's about a year old now. And that's an amazing breakthrough that most people I think don't even really understand how important that's going to be for designing new drugs and understanding the body. If they had had years ago I probably would have stuck with it, but really interesting work where AI is applied to so many different things.
Emm:
It's going to be beyond drugs, beyond proteins. AI might be applied, don't be scared. AI might be applied to create new organisms.
And it's going to start with microorganisms. And then we might go into like multicellular organisms. And that's what I was doing back then in 2010. But we were like basically randomly trying to select, So we were trying to guide evolution, basically, with robotics and specific medium to select variants. And which was, I mean, we got some results to create new variants of yeast for producing bio-bio compounds and biochemicals. But I think with Stanii now, to design new genes, new proteins, you could create new organisms, which sounds a bit postient, right?
Jon:
It sounds very exciting. I think we have to be careful about certain things for sure, obviously. But I guess that's an important takeaway for everybody watching this, which is there are so many applications of generative AI. It's really about bringing new things into existence, speaking things into existence. We've talked about games, people have looked at art, but it applies to an incredible number of things. It's about unlocking biology and figuring out new It includes really all forms of science will now be impacted by AI. AI will become a new tool for probing reality.
Emm:
What about climate change, you know, like using generative AI in some ways to better understand and then maybe fix the issues caused by climate change.
I mean, a human impact on ecosystems overall. So it's not just about like fancy pictures in a game or weird like chats that you can have with the machine. I think it's way deeper than that. And as people say, it's a platform shift that's happening right now. How exciting is that? You know? way deeper and way more concrete than blockchain even though blockchain has some pretty interesting aspects when I tweeted about it recently but it feels like this will go way further faster.
Jon:
Yeah. Well, when we started, I talked about how the industry might change with technologies like scenario, making it easier to create things. We just talked a little bit about how science is going to change. How is the world going to change? How is society going to change? Like, what are we humans going to be doing down the road? Let's dream for a moment. Like, what's the world look like five, 10, 20 years out as AI absorbs more and more of this?
Emm:
So I was struck by one tweet recently, and I forgot, unfortunately, I will try to find it. One person saying, showing that GPT-4 was passing every exam, every major exam, physics, law, medical exams, and so on. And not just passing. They were among the best.
And the comment of the tweet was, the cost of intelligence is quickly dropping to zero. be extremely expensive to develop intelligence, like training a human, like a physician, like even like a lawyer, everything. It takes time. It takes 20 years of the human growing, and then five more years of university and college, and then way more years of experience. And now with these systems, we have somewhat, somewhat, because it's not quite a human, of course, but we have a way to approach their level of intelligence for 20 bucks a month. I mean, think of the implication. It's bringing intelligence to every, to the 6 billion, or 8 billion people on the planet now for 20 bucks a month, or maybe less. What are the implications? Hey, we should see an explosion of new technologies. Oh, and not just technologies, by the way, art as well.
I mean, supposedly it should also increase creativity, not just on the tech side, but maybe the art and culture side as well. So, and that would like, We are entering an exponential curve which can be scary and exciting at the same time. That's what I think.
Jon:
It's on the art and culture side, it feels like we could be at the threshold of like entirely new art forms really we're just starting to see it now with with some of this new stuff but like AI rather art always sort of goes through many revolutions over time, like photography was a huge future impacted by this technology or maybe you have thoughts about it as you thought.
Emm:
I mean, I'm an optimistic person, so I always expect the better.
Have you seen how people have tried to also, you know, kind of detects brain waves that could be used to reproduce arts or images from just the brain waves?
And so that sort of new art could be developed where like everything you see is actually only specifically developed for you based on the signals the machine can get from you. Maybe not just brain waves but I don't know all kinds of you know signals they could identify even visually.
Emm:
Who knows like it's a pretty interesting time.
Jon:
Yes, there was this research came out quite recently and it showed that you could put someone in an MRI, scan their brain, and there might have been a lot of training and stuff going on. Who cares? Like the fact that you could look at these patterns that were present in someone's brain and be able to create a version of what they were looking at on a computer screen, we're gonna put a lot of links and show notes for everybody to look at after this, because all of this research is super interesting. But you were mentioning dreams, maybe you could record some of that. This is probably not super close, but you could
Jon:
recording these surrealistic scenes that you encountered while dreaming, and maybe that's a new form of art.
Emm:
Yeah, that is super cool. It's science. It's definitely science fiction movie material. But now with JNEI and with the latest news about, like, yeah, this MRI, guys, it feels like we're not too far from it. There should be a Hollywood movie first. Then we'll do the tech, like always. And GPT-4 should write the scenario. You know what? I'm going to give it a try later tonight.
Jon:
Yeah, and science fiction books. They're science fiction books tend to be ahead of everybody.
Emm:
I mean, maybe it's like a very easy answer, but I've always been fascinated by the matrix. Like, come on, like, can you believe the matrix was really what was like 1998, right?
…the late 90s, right? With this parallel word and I think that that's a film that the movie that really struck me in sci-fi. And then, I don't know, other favorite sci-fi movies. I think minority report was pretty cool. Maybe dark for sure. I think it was more, you know
Jon:
Now those are both pretty dystopian.
Emm:
I think it was more, The technology you know what? that's right. I mean, the matrix, I believe, was not the tech, but more like the concept of what is the matrix supposed to be. Minority report was interesting for the, like, the whatever screen they had to, you know, quickly
Jon:
Yeah, the augmented reality.
Emm:
interact with many visuals. Yes, so when I first tried my first VR headset, I'm like, okay, this is minority report. Shit.
Jon:
So what's a day in the life for you like? Like what gets you up in the morning? What's your morning? What's your day? How does that look?
Emm:
All right. So I have a team in the US. I'm based in the US in San Francisco.
I have a team in France as well in Europe. So the mornings are usually pretty dense because I wake up with already a bunch of notifications on my stack and discord. And so I spend usually I wake up at 6. I spend a few hours trying to catch up with the news of the night, drop the kids to school, and then by noon, that's where I stop doing emails and answering all these notifications. And I try to do some more deep thinking or focus time in the afternoon. I'm talking to studios every day, literally two, three studios a day, today on average, from all sizes. When it's the most exciting part, I think, for an entrepreneur to confront your idea and your vision to a market. Sometimes it's scary. Sometimes you're afraid that you might just be not interesting at all. But so far, people are very interested, and we get some good traction, especially from the GDC event coming. And it's the Bay Area. So you come up pretty early at home, around 5, 5, 6. I usually spend time with the kids. I play a bit as much as I can. And then I work again until I can at least talk to my team around 11 PM, midnight, 1 AM, and I go to bed. So they're pretty dense days. And then I try to disconnect at least one day over the weekend to go fishing or swimming or hiking in a beautiful bay area.
So, pretty dense.
Jon:
I get out into the world myself. I've climbed some pretty big mountains out in the world. So I get it. So it's good to go on to the outerverse in addition to being in the metaverse all day
Emm:
I go to the ocean, the ocean verse, like at least you're sure you have no connectivity, no air pods, nothing's working. You just dive with the fish and I love it a lot.
Jon:
It's interesting how many things we have in common. And also, even sort of, although our specific problems that we're solving are so different, like for me, the problem I looked out in the world and saw was like cloud native infrastructure for games is just so complicated. Like I don't think any game company should have to build these things. So I was like, if we could just let them push a button, dream things into existence, and all the multiplayer functionality comes to life, that's great. But then you did sort of the part immediately is adjacent to that which is okay. Well, you have to bring that life that world to life with imagery with You know objects art as you said music levels all the content of that world So it feels like these two things converging together is gonna bring about a huge era and creativity Where you can I've used this phrase before that the direct from imagination era I just want to get to the point where if you can dream it up you can get it onto the screen fast
Emm:
Yeah, and it's great to be a creative. It's great to be a creator, but I feel it's very rewarding to develop technology or a tool for creators.
It's like you see these guys with amazing ideas bringing them to life using your products and others. But I think it's really, really, really rewarding.
Emm:
That's why I love the post. When I look up at the Twitter notification every day, I'm like, okay, who has made some cool stuff with scenario today? beyond how many daily active users or retention rates for the months. Like I'm looking at what's the cool image on the Discord. That's the energy of the day, usually. I'm looking at what's the cool image on the Discord. That's the energy of the day, usually. I'm looking at what's the cool image on the Discord. That's the energy of the day, usually. I'm looking at what's the cool image on the Discord. That's the energy of the day, usually. I'm looking at what's the cool image on the Discord. That's the energy of the day, usually.
Jon:
Yeah, so don't fear these things. Like, like really learn about them. I think people should be encouraged by it really, like, like, go to the edge of your capabilities and then use these things to help you go even further and then multiply yourself. I think it's going to be a very, very exciting time to be a creator of any kind.
Emm:
And by the way, even like if you look at a very cold, if you look at this, every major technology was built on the fringe of what's acceptable, legal, good, bad. Like it was always like a blur. Every big technology was built this way. So this should be no exception. And of course, the challenges that we still face, such as IP, copyright, training databases, and so on, they need to be discussed. that needs to be some debate around ethics. I feel people, generally speaking, in industry recognize that. And I want to fix it.
…it's specifically talking about opting out specific artists. They are talking about reducing the size of the database, making smaller models so you can control the database better and so on. So the challenges will be solved. It's a matter of maybe a month, maybe a few months, few years. It will be solved. So there is no way to oppose the technology that the problems will find some solutions.
Jon:
Exactly. And I think Emad has said this, but it reflects my own opinion too, which is, yes, have the conversation, not everything, we're not going to get everything right. Obviously, technology always sort of explores these dark edges, but have the conversation in public, you know, in the open, don't do it behind closed doors and then have a few big companies decide all the rules that everyone has to live for. I mean, to me, that's just regulatory capture. And I'd rather see startups and anybody who wants to do that. experiment with these things, have a chance to go and build things and see what we can bring to the world.
Emm:
which is the case today. Thanks to, thanks to Hugging Face, thanks to Stability Fusion, to many other people posting their stuff on GitHub. How amazing is that?
Jon:
Great. Great place for us to conclude. And this has been a really amazing conversation. I hope people are as inspired by this as I am. I think we're going into an incredible future here. And again, don't fear it. Embrace it. Learn about it. See how you can apply it to your life, to your creativity, to everything that you do. Would you like to share any final comments before we sign off here?
Emm:
But yeah, I think that's the greatest conclusion. Even if you fear it, have some courage to try it. And I'm pretty sure the vast majority of people that tried for at least a long enough, a few days, a few hours, even a few hours, should at least say, huh. Yeah, maybe I should not just go all against it right away because I read a few articles that were already on one side. And there is this thing that says AI won't take your jobs, but maybe people using AI will take your jobs. I think that's fairly right. Like if we look at it like a very good, so another code look like use tools, otherwise people using more efficient tools might really compete hard with you.
Jon:
One could have said the same thing about the internet 20 years ago, like internet became a tool that everybody learned to use to amplify whatever work they were already doing. This is probably going to be even bigger than that. It's enabled by the internet, actually all the data collection and networks that we have are what actually enabled AI to become so well trained, but use it like, and don't fear it. We were talking science fiction earlier. You know, fear is the mind killer. That's a good dune quote. So learn how to learn learn how to succeed in this and grow in this new era we're going to.
Jon:
Thank you so much, M. This has been an amazing conversation.
Emm:
Thanks John, see you soon.