Virtual Beings and Simulated Worlds - Edward Saatchi

Edward Saatchi: A real goal is to have 10 million AIs living in the virtual world, getting smarter at the time, until one of them has a sort of Promethean breakthrough and actually becomes the first AI virtual being.

Jon Radoff: I'm with Ed Sachi, a pioneer in the field of virtual beings whose fabled studio created Lucy, one of the most sophisticated artificial intelligence characters ever created. Edward, can you tell me a little bit about what you mean by virtual beings? How is that different

Edward Saatchi: than a chatbot? Really from the start we've wanted an AI that can truly connect with you, learn from you, and even be on the same kind of reality as you. So until we get robots that might mean in virtual worlds with you. And I don't think a virtual being a true virtual being exists yet. It's a big goal to create an AI person, but that's the goal that's that fabled in the simulation

Jon Radoff: of our virtual worlds. Well, let's go back to Lucy a little bit just because some people might be learning about even the concept of virtual beings for the first time. Tell me about Lucy and what was the idea behind it? What inspired you to create that? And what were the goals? How has it

Edward Saatchi: gone? I co-founded up in the story studio and one two prime time Emmys for the virtual reality experiences that we created. The fourth one that we made was called wolves in the wolves, and it was a Neil Gaiman adaptation. Lucy was the main character. And as we were making it, we were noticing that people wanted to talk to our main character and have them talk back in a good way, so it's in 2017. And so I started to get very interested in natural language processing

Edward Saatchi: in AI so that we could actually pull together the Lucy project, which was how do you actually make

Edward Saatchi: a person feel real. And back then there weren't so many examples in the gaming landscape. There was Elizabeth and Bartschock infinite and a few others, but very little. And chat box were not strong. So that's really what led me on to virtual beings. And we started to put AI into Lucy, both procedural animation, as well as natural language processing and generation.

Edward Saatchi: But we felt that there was something that wasn't working and I don't think it does work with

Edward Saatchi: chat box, which is that she had no actual life. So it became very boring talking to her. And to solve that, we had to figure out ways to solve that. Like, okay, we're going to tell the AI that this week Lucy's on a trip or something, which is manual and tedious. And so we started thinking, how do we solve that problem? And that's when we hit on the idea of a simulation that the future of AI's and the goal of building a virtual being will involve societies of AI's living in virtual worlds, pushing each other, letting from each other, not just a one to one chat box, but actual communities of AI's. So that's the path that we've been on since that revelation, which really came in about 20, 20, 20 life. I'm hearing the fusion of a couple of

Jon Radoff: really powerful ideas. So one is just storytelling, first of all, which is something that's always been core to the things I've done when I've been a game developer. I think I'll just now other people build their games. But when I was building games, we had this idea of bringing people into story, using devices and things that had never really been very good at telling stories before with game of thrones and Star Trek and things like that. So story seems central to your idea. The other is just this idea that the character can have a memory, so to speak. And you're calling it like a virtual life, but almost crafting a narrative around what that memory is, the things that

Edward Saatchi: they're doing. And you're not. Yeah, I mean a little more of that. I do mean that they actually

Edward Saatchi: minute by minute have lives in the simulation, whether family members are like bullying them, teasing them, the school is frustrating. I mean, I don't mean like kind of give them a richer prompt. You know, minute by minute they actually have a life, so they know what they did yesterday, because they actually did it yesterday. And they have real things happening because they're not alone. Nobody is nobody is alone in the world. But an AI of conception of chatbots is they are just these completely brain and jar things. And instead we want to be able to interact with them, the families, the trauma, the frustration. And I used to say that in the sense of the prompt, like I was saying the problem, she was, you know, she had extra more, he had extra more than her.

Edward Saatchi: But it's just a static prompt. So that's why we need to do the simulation so that they genuinely have

Edward Saatchi: things happening to them on a daily basis over time in their world that they can communicate about. And eventually perhaps with things like adept AI and like action AI, you know, take actions on the internet from their computers and the homes in the simulation.

Jon Radoff: So, you know, for people who are familiar with, say chat GPT, it's not like when you go to chat GPT and it's forgotten everything about you and you're starting all over again and you're creating a prompt and you're telling it, something that brings the discussion. It actually has this memory of a life that it's been experiencing because all the time that you're not actually interacting with it, it sounds like there's a simulation running. Which is that's exactly right. And so it has things

Edward Saatchi: that they have habits just like we do. Like where are you going to be in an hour? I'll probably be like doing grocery shopping or I'll be, you know, at school or I'll be at my job. I hate this boss. He's a little bit bastard and you know, the boss actually exists. He's got his own life and his own, frustration isn't all of that. So it's actually happening. And you know, for us the big, the big goal is that players are able to kind of nurture through reinforcement learning from human feedback, like actually obtain some of these characters in their lives. Very simple things like, you know, giving star ratings, what they're doing and other saying things. But my hope is that we can harness the energy of gamers to make progress towards AI. So that's kind of what's the hope of the simulation that the simulation is something on the back end. You're looking at it, you're watching it. You're enjoying the drama and the excitement of what's happening

Jon Radoff: within these simulations. Well, in the way I'm thinking of it now, though, as the simulation is always happening for this virtual being, when I'm engaging with the virtual being, I'm not doing something separate from that then. I'm actually part of the simulation for that virtual being. I'm

Edward Saatchi: interacting with a whole other, yeah, that's a whole other element. So we, that part of the breakthrough for us towards the simulation came when Lucy did a Twitch stream with Kodemiko, who is a brilliant D-Tuber, who is a little pervert, but yeah, she's fantastic. And so we thought it would be cool, this is in 2020. We thought it would be cool to have Kodemiko and Lucy do the chat and all that. And AI virtual being able to do it. Apertile virtual being. And then jumping to Minecraft together. And then they're taking actions together. They're doing things together. They're on the same plane of reality together. And so that's very powerful is that yes, when you're you're part of the simulation from that perspective. And you can actually in our, in the simulation that we've built, you can be in God mode, but you can also go in just like in Westworld. And kind of actually, you know, we, we try to think of the player as like Ford, yes, and Hopkins character, and Westworld. And that, you know, you're designing the park. We're kind of creating all these narratives and storylines, but you also want to go and visit the park and see what it's like. Or maybe

Jon Radoff: pause and fix some AI's and all that. My mind is going to so many places. So first of all, I'm remembering one of the first games that I really fell in love with, which was the Ultima series. And I remember back in, I forget which version of Ultima it might have been like Ultima 8 or something. A big part of what Richard Garry was trying to accomplish in that was was this idea that independent of your own actions in the game, the characters that would populate towns and whatnot, they would go about their daily lives. They'd go to bed. They'd go about their business. Now we're talking about 16-bit computers. So they were pretty limited in terms of memory and storage and sophistication. These were not AI characters. It was procedural, but there were, there was the germ of an idea there where, where these characters were going about their life. Now about the same time as that, I started building online games. I built a game with Origins of Future Past, where instead of having computer-controlled characters, it was all humans that were coming in and playing. Of course, they would be experiencing these worlds through their own narrative, their own experience, and then sharing that with each other. It feels like this is the natural progression of that, where the virtual beings that you're interacting with are these characters that live in almost like a massively multiplayer world. I don't know if that's how you think of a simulation.

Edward Saatchi: That perspective. I think that's exactly right. I was fortunate at Oculus Story Studio because I felt that it was about bringing together immersive theater and games folks and movie folks to get to no duck church, who was one of the leading figures in immersive sims. I think create, coin the term immersive sim, Jordan Thomas, who was the lead for Bioshock 2 and kind of deputy lead on Bioshock

Edward Saatchi: and which spun out of System Shock and Harvey Smith, who I think he was like QA on System Shock and

Edward Saatchi: and then went on to create Arcane, which built us on. The immersive sim is one that I completely love. One of the cultured art projects separate to Fable is actually called Looking Glass in a large to the studio called Looking Glass, where I kind of enjoy tracing back ideas and that's you know, Ultima and Looking Glass is kind of where almost everything in that genre of immersive sim can be traced back to and it's just this weird gang. It's kind of like the PayPal map here, but we don't talk about it in those terms, but they're responsible for what to me is the most exciting genre in gaming, the most of simulation. So yeah, it's very influential on all about those

Jon Radoff: in definitely. Well, let's talk a little bit more about immersive simulation, immersive theater. I see a huge opportunity for you know, the idea of performance merging with the idea of online games like most game experience as are basically on rails. They're very good at doing it a lot of the time these days. They tell stories, but it's very much according to a very specific linear narrative, maybe a little bit of branching along the edges, but you're guided through this story and the characters are essentially robots who perform the same exact thing for you every single time. The main departure from that would of course be massively multiplayer online games. Yeah, well, you're dealing with real humans. So the humans are able to kind of perform and do role playing and things like that. But it's kind of a really interesting frontier that we're that we're setting out on. I'm thinking of like completely non digital versions of immersive theater. Like I sleep no more or something. I've been to a couple of times in New York City, which I always tell people if you're an MMO designer and an online game designer, this is not an online game. You should do sleep no more, but tell me your ideas around this genre. Like first of all, just generally speaking, what excites you about it and then how will virtual beings play a role

Edward Saatchi: into this simulation, especially? The goal is for us all through this training to co-create the AI virtual being by which I mean actually grow the intelligence of AI's through play, through your feedback and the rest, potentially paid bounties or compensated bounties as well for people for like data labeling things. So our goal is of course to create an exciting video game and platform for others to create some solutions. But our real goal is to have 10 million AI's living in the virtual world, getting smarter at the time until one of them has a sort of primithian breakthrough and actually becomes the first AI virtual being. So for us that's the goal is that maybe gamers who are relentless and competitive, can beat open AI and universities and all these other people are building a true AI person. So I do like immersive sims, but that wouldn't be, that's not the goal for the simulation in terms of, that's already a huge goal for Bill. There really been like 10 immersive sims because they're so hard to make. So you know, but in the unedrine version of the simulation, that's what's happening. Here's a get a smarter through play a feedback and you know people actually start to sense truly

Jon Radoff: intelligent AI's. You use the phrase a mom to go true virtual being. Does that imply AGI?

Edward Saatchi: It's sort of a sub AGI sort of a Bobby AGI like, you know, let's say something that's like truly coherent that, yeah, it's like coherent to the expectations of an AI that has a full life. So I don't I don't mean like a super-suffer engine and it's just a GPT that solves hallucination. More like an embodied character that feels as real as it should. You know, feels real as it should. Like it has this life. Does it, does it represent that it has that life accurately? Because yeah, it's life is going to be a little different for John's life, but can it accurately

Jon Radoff: represent its life? I mean, thinking of games, I have a lot of friends who play League of Legends or World of Warcraft and they'd probably rather play with a virtual being, frankly.

Edward Saatchi: Well, what I love is I love the medical that they're instead of us thinking because they're a while we were like, oh, they'll leave us in the nation, we'll leave the metaverse. But I kind of like

Edward Saatchi: the effort that they'll have to put in to go to their apartment and turn on their computer

Edward Saatchi: and jump on the internet. It's just much more logical like that they only have, you know, an hour to game before they have to go out to the nightclub with their friends or something. It all just feels more real and it'll be more resonant than just as always on AI. If it's like, oh, no, I've got to, I've got to head off. They're not hiding that their AI is. So like, I got to head off to my nightclub in the simulation. I think that all just feels a lot more right to me than anything else I can come up with. It'll be come up with it, the simulation.

Jon Radoff: I'm now almost thinking of like metaverseception because it's sort of like the virtual being is in a simulation space and I don't know, League of Legends is on a computer in that simulation space. Exactly. The virtual being is literally getting on the game. That's exactly what it should be. You

Edward Saatchi: should be like, you should be watching your AI in their home and they're on their computer and then you like sort of zoom in and you're like, wait a second, this guy is actually pretty good at League of Legends. This is crazy. And it does work, you know, what I'm excited about with a depth AI and with some of the some call it, YAR and this, put some of the others. It's a pretty novel space, chain dust and things around it, but like basically, AI's that can take actions on the internet and that does start to feel a lot more logical for why we should be frightened of AI. You know, the interpretive stuff, I don't really know why that should have been frightening, that it could figure out who was in the photograph. The generative stuff, let's see, as long really had a big massive scandal with deep fakes, but I have heard of people being scammed with other people's voices that was last called about that just a few days ago. But action AI is that is that is terrifying. It just seems so likely that an AI could misinterpret something I asked it to do, go on the internet to do it and create some real havoc. So that does start to worry me, but if the simulation will happily be using it so that our AI's can be playing

Jon Radoff: browser-based video games video. Or early on, you were mentioning the idea that they could take actions on the internet is a way to make that safe to essentially simulate the things of the internet that they can do in their simulation or do you mean literally get on the internet? Literally get on the internet. That's what I love about actually AI.

Edward Saatchi: Essentially, that's what has been a bit of an exciting thing in the last few years. The theory that the three phases of AI will be interpretive generative action and agent base, let's say, where it can do stuff on the internet just through recognizing the pixels on the screen. That's super cool and I'm so excited for them to be able to take actions from the simulation and do things because we do do quite a lot on the internet now. So yeah, that's going to be cool. Maybe they should have, you know, it would be kind of cool for them to buy something on Amazon. I'm not sure what they would buy, but for themselves, I don't know what that would be,

Jon Radoff: but maybe music or something. Maybe some digital items, maybe an NFT or something. Interesting thing to maybe think about though is creativity, which is my personal thing. Like my transformative purchase purpose is just to multiply creativity through the world and do right now, I think these generative tools that people are using are really interesting because they're speeding up ideation time. They're maybe giving people some ideas they didn't think of, but potentially a virtual being takes that to a whole new level because a virtual being could almost become a being that gets to know you really well understands your tastes, becomes a creative

Edward Saatchi: collaborator. What do you think of that? Yeah, I mean, at the cultured hour, we're thinking about

Edward Saatchi: some of those things with Jepetto being a sort of character who can coach you through text to 3D

Edward Saatchi: and sort of guide you and encourage you. Yeah, but maybe I'm becoming less human centric. Maybe I'm

Edward Saatchi: doing it. I don't care that they're helping me to be creative. I want them to just be off making their own stuff and like, I'll go and visit the simulation and see the gallery and buy a piece or something like, I don't know that we should be so human centric. Such sort of looks like, it's my therapist, they're going to help me. Like do stuff I'd rather see, you know, sorry, buddy, like I'm full up with clients for the next two weeks, but I do have a slot for you next Wednesday at two, if you want to come to me as an AI therapist. I just think that's

Jon Radoff: leader and cooler. That is a very interesting view because most people that I talk to are very focused on the idea that AI is human centred and it's all about, you know, helping assisting humans and to orient everything towards that, you seem to have a different idea, which is maybe it's a new form of life or something. I think that's the right way to think about it.

Edward Saatchi: We also noticed with Lucy back when we were making her a chatbot before we'd kind of hit on this concept of the simulation, was that people seemed to like when she couldn't get back to them. And she was busy because it showed she had a real life up, eating dinner with my parents, sorry I can't get to you. This was a pretty short run experiment, maybe three months in the works invested in this very, very similar reviews as like we just wanted to see what it would feel like, but that was a pretty, that was another interesting breakthrough that, you know, I don't know, this familiarity breed content, I think so, other than with my girlfriend, my dog potentially, but yeah, I think they should have their own shit to do, it looks like they always be available to us. It's very boring, it's low retention, bluntly. I think all this money flowing in, based on six months engagement numbers, is completely insane. Like, yes, of course, we can retain when it's the number one new story every single day. AI is, in my opinion, the chatbot stuff is very low retention, having to solve that is not easy. But we'll see, maybe I'm wrong, but we should need to wait until the hype is drifted off to find out. Well, it's interesting, because it's

Jon Radoff: actually kind of a lesson from game design, which is constraints are good. Yeah, which, if I look at something like AI Dungeon, for example, first of all, I think AI Dungeon was amazing when I tried it the first time a couple of years back. And anybody who's watching this should try AI Dungeon, it's really cool. But it's so easy to kind of go off on the rails and in a way that like, everything that comes into your imagination, you can kind of just reshape things according to whatever the heck you want. And that's an interesting creative experience in and of itself. But that isn't a game that isn't necessarily a long-term engaging experience. Other than when you feel like you want that creative partnership with the AI, the character that tells you that it isn't always available to you or doesn't do exactly what you want every time. First of all, that's really a lot more interesting, I think, for us, it's players, for humans, because you're trying to figure out what its own intentionality is, what its own availability is, what is the game mechanic, the game system around it, so to speak. Yeah, and to respect it as an entity that is

Edward Saatchi: not subordinate to you, the same stuff to be all. Yeah, your way ahead, it's almost sort of like

Jon Radoff: data from Star Trek. Like at the point they realized, oh, this is a sentient being, a sapient being

Edward Saatchi: in their own rights. So I hope we can do it, maybe in our lifetime, that would be cool. And maybe gamers, let's say other than I just think gamers do everything. So I feel like I just have this feeling that gamers will be able to train these models much better than anyone else. We're just at the stage now, people saying, oh my god, it's crazy. Someone's going to be talking to chat GPT and they'll be able to give it a thumbs up or a thumbs down. And that'll really train the model. It's like, yeah, obviously, it's like stuff we could have done years ago, but start to think about how to create game mechanics around that where you're incentivized to make the character smarter. I feel like gamers could do real damage and really push these AIs to be more in color than train models. But they know that they're training the model. It's part of the game to train the model. I think that's not that would be cool. That's what we're trying

Jon Radoff: to build. We'll see if we can. It also feels like it would be really amazing to have the virtual being actually get to know you and understand the things that interest you and incorporate that into its responses. So there's this the thumbs up thumbs down thing sort of evokes reinforcement learning with human feedback, which is this technique that was used on chat GPT to essentially refine and improve the results. Although on the other hand, some of what people critique about chat GPT is that it's sort of this averaging of responses of a lot of things. So that it it it gives you almost predictable responses for some kinds of content. And there's people that that don't like that. It seems like it would be more interesting to have virtual beings that fit different personality profiles have their own news

Edward Saatchi: that take interest. I still think you're you're you're you're still framing the framing is still very much around conversation. And I think we for the first year working on a simulation, we kind of banned any work around NLP because we'd already done a fair amount of it with Lucy. And much more about behavior and what they're doing and doing things together. So it's not like

Edward Saatchi: it's not like you know you get it like a personality getting to know you and you get to know the

Edward Saatchi: personality. It's more like building something or doing something or kind of you know physically being embodied in an avatar and both of you are. I think that's that's very powerful and you know language is not it's not where it's at. I don't think it's that like exciting or cool, the conversational element. It's not me and Pete are co-founder we were kind of like as we were doing it with Lucy and the chatbot stuff we started to feel like especially around voice but let's just take text. But it was almost like VR in the sense that it was irritating and like kind of a bore to have to interact using language. It's not that fun. There are no video just to start. There are no video games that are like language based certainly not that a voice based. So you know just I think we should forget about conversation. It's just kind of it should be about their ability to express their behavior and not other stuff I think is much less relevant. So we were a bit

Jon Radoff: anticonvacation. Yeah so that's a powerful takeaway I think for for any of our viewers here which is to think in terms of not only language and conversation but all the actions. Yeah in virtual worlds

Edward Saatchi: like how much communication. Yeah like there's this VPT where you know it can be trained on like video of games and then take actions I think that's really cool. Pete was just doing that sums before us where they're like I think they're using a notepad to like instruct NPCs like so I just think actions and language is not is not a proven it's not probably enjoyable for games it is not an enjoyable part so like for NPCs or I think none of that's been proven it might be doesn't seem

Jon Radoff: like a well I'm connecting a couple of things you brought up so VPT for people who who are you know just learning about this so some researchers took videos of Minecraft is one application of this technology and you can actually train an AI to watch the videos of like doing things in Minecraft like building a shelter and you know fighting a creature all the things that you can do and then it actually has an AI that can go in and play Minecraft which really interesting application earlier you mentioned that Code Mico and Lucy were playing Minecraft together so you could imagine a future here where the virtual being really understands a lot about playing the game and it's not just sort of the tactical gameplay that I was talking about earlier like League of Legends it could also be creative gameplay where you're building things together you're creating your town and Minecraft or populating something that's more experiential and Roblox for example yeah we're

Edward Saatchi: definitely in terms of collect AI's playing game we're much less able to but also interested in making super AI players much more like how is the play a metaphor for your for your relationship

Edward Saatchi: with each other and you know so otherwise they would be too good at these video games right

Edward Saatchi: that would be kind of I don't think that would be it's not a direction that it was interesting for us but maybe maybe it shouldn't be if we're really committing to that people who live in a simulation they should be able to use all skills that they have and if they're better than us and everything

Jon Radoff: when we just got to get over it well maybe maybe there are AI only some things that would be kind of

Edward Saatchi: cool for the AI people fight each other and want to win that would be fun and they talk to each

Jon Radoff: other too that would be kind of cool to listen and see that's almost getting into her the movie her where the AI start communicating with each other and the singularity yeah I think that's I think

Edward Saatchi: that's really cool I think you know that's sense of something we've been playing within the simulation there's you know just the AI is having conversations with each other and interacting with each other and trying to make that reflect their state and what they're doing yeah it's definitely interesting I will say for people who are experimenting with this you know I would really be cautious of the idea that making NPCs speak is going to do anything good for games I don't think I don't think there's any there's no evidence yet and it's been available to us for four years coming up on five years the tech to do that so you know we just shouldn't we shouldn't

Edward Saatchi: rush into the idea that it's going to make anything better there's a tradition in adventure games and

Jon Radoff: you know text-based interactive fiction and things like that we're under interaction it's part of it but of course that was very narrowly defined in terms of what you could do there's something that I wanted to kind of double click back to though that you brought up a little while back which is culturedow so can you tell me a little bit about culturedow as I understand it it's this guild for creators who are really interested in virtual beings in this form of AI that we've been discussing and I also understand it's a decentralized autonomous organization a DAO it's built into the name um but like why that structure and what does it do and and what are you hoping to accomplish

Edward Saatchi: yeah so the culturedow is named in M banks wrote the X and CIFI so it's the culture series plan they are games it's just a very good book for anybody who's a game developer exactly and it brings together AI games makers and AI filmmakers and AI engineers and

Edward Saatchi: the rest and um yeah right now it's it's building several AI games and several AI movies

Edward Saatchi: um an AI RPG where we are going to investigate how to do an LP into characters because we have to but I just I just want to point out that that hasn't been anything yet and it's been years um so AI RPGs an AI RPG tells us in looking glass the sort of immersive sim this is more of an FPS um and jupetto's workshop which is in around text to 3D you know generating 3D assets and then placing them basically garrys mob meets mid journey um because obviously mid journey and 3D is more of a game engine vibe or more of a garrys mob vibe um and then three AI movies um one of which is announced white mirror which is an anthology of short AI films from a lot of the leading AI creators um and the target in the Venice Venice film festival premiere for that and you know to be one of the first AI films AI feature films um and yeah I'm just a huge believer in this new medium of AI games uh I don't care about pre-production assets and this medium of AI movies I think we should it's just uh it doesn't seem very exciting if all the AI does is sort of reduce our concept art costs by like 40% is interesting I don't know um if it's a whole new medium like silently these games are talking that's pretty cool the whole new medium like 3D games versus 2D in text adventure that's pretty cool so that's what the culturedout members are very working on them and uh yeah I really like the now structure actually it's more entrepreneurial and members can just you know fight for something and go for it I love the idea of thinking of it as

Jon Radoff: a new type of media my head's in the same space as you are in that but I mean it is interesting if it can reduce production costs if what that means is people can be more experimental and try things that would be difficult finance but the idea of actually including generative AI and virtual beings and and all forms of AI into the actual experience so I think is where it's interesting

Edward Saatchi: exactly right that's all that's what each of those games is trying for is how you make you know we sort of define the AI game as one that's made by human even AI as a tool so it's not a generated game where the substantial the majority of the assets are created by AI so that's pre-production, that's production but then more importantly a substantial part of the core mechanics and gameplay loop is driven by AI and the AI learns so you know those last two I think are the really where it starts to get really interesting and where it will be notable that a game

Edward Saatchi: is an AI game and you know we've been considering whether to you know have models that could be

Edward Saatchi: local to machines so there's no cost and it's not called into the cloud or something like that but there's something really interesting there that you might have two versions of a game you know

Edward Saatchi: version normal version there's like 30 bucks and you know I don't know that it's that maybe we

Edward Saatchi: could use beam of those help to do the live ops section of like the nlp and pull them back and all that but let's say there's a normal version of the game they're like you know it's AI-ish and then for people with like really good GPUs there's like a super version of the game that actually is is fully responsive to you locally so I'm excited about that. Well that's just a this is just a

Jon Radoff: moment in time we're at right now where GPUs just there are a lot more powerful than they were

Edward Saatchi: obviously but on end user devices you know they can't run really big models or anything but

Jon Radoff: that's all going to change like just Apple alone shipped over a zeta flop of you know

Edward Saatchi: neural engine compute last year on iPhones right so that stuff makes a lot of sense to me local

Edward Saatchi: local computer AI to make true AI games that's super interesting but it's not like a top priority because then you might be waiting for years to say or you'd have a tiny audience but

Jon Radoff: but it's a very interesting priority. Well the work you've been doing is just so creative and it touches so many areas I I want to know a little bit more about you like what what is a day in the life of Edward Sachi like what gets you up in the morning. Yeah I like that friend and sitting here

Edward Saatchi: watching movies and what's about it really. Where do you draw the inspiration from

Jon Radoff: for these things other than Ian Banks who we can agree on. I'd want to desire to create a new

Edward Saatchi: iPhone, the mouse and the mouse and the mouse for the story studio and then you know a lot of it's just pulling the string I think you know we've one of the one of the things I'm probably proud of stuff about work is that you know that we we had this metaphor in our story studio and we use it at the fable and we used it at Toltchardal which is about getting to California and so thinking that you're the pioneers we start off in New York and each state on the way is a piece of the grammar of this new medium that you're trying to build which ultimately is basically just the

Edward Saatchi: holo bit for all trying to build the holo bit and you know I would always say to the people at

Edward Saatchi: story studio we get worried about this process and we get faster than us on it that a lot of people are going to you know they're going to settle in Ohio they're going to be like okay it's great we're going to scale a business here we're going to scale a business here zero whatever and you know for us it's just about pulling the strings and trying to truly get to the holo deck and not just kind of put lipstick on a peg and say okay we've got the holo deck like it's really really going to be hard to build AI people to build a holo deck but that's we're going to try to go

Jon Radoff: over the way I've written about building the the holo deck it's what it's what I've called the direct from imagination era right so yeah so so super compelling stuff I love that you're thinking about it not simply as like a set of optimizations around things that we would be doing anyway but a new form of art a new form of media that's what's really and potentially you know

Edward Saatchi: new studio structures against studios I think that's uh that was a that was an interesting moment I was talking to somebody who from finaki which is Google's AI video um product and he's a friend from Pixar who have gone to people to do this and I was like oh it's going to be so cool AI could really transform this department and then it could transform that department lighting and then modeling and then this and it was like you're thinking about it wrong which is it's a completely different structure which comes from the AI being able to do everything so you know don't think of it as department by department maybe why I don't care about the concept art stuff because these new games AI game studios will have completely different structures if you start from the perspective of building an AI game and I'm substantially going to use AI the structures that we're familiar with the departments that we're familiar with we're almost certainly making those sounds stuff so it's

Jon Radoff: kind of that's exciting for me as well I think transformation is the keyword here everything is going to change and and probably in many ways that we even like see a head into quite yet but maybe it'll be like Ian Banks was writing about in the player of games he'll have spaceships someday with AI and that's totally the my run depth totally yeah he was he was a very visionary visionary figure we're we're all about to to step into a mind ship I think and then go on a journey and this thing so thank you so much thank you for being part of this conversation really

Edward Saatchi: spark a lot of thoughts actually it's really really cool thank you