Game Development Livestream

Jon Radoff: All right, everybody. Welcome back to episode number three of the game development live stream. This is going to be a really, really cool topic today. It is going to be well worth an hour of your life if you are anybody in game design and development. So we're going to be talking about game engines, procedural content creation. We might talk about generative stuff and the difference between the two. But it's going to be really great if you're thinking at all about a new game engine for your project or switching engines or want to complain about the engine you're in and think it should be better. We'll talk about that. But let me introduce my co-hosts and guests today. So first, Paul, my co-host.

Guest: Howdy, everyone. I'm Paul. I've been making games for 25 years now with a lot of different engines. And surprisingly, a lot of emphasis on doing things procedurally. So I build myself as a system designer. So I'm really excited to give this topic today.

Jon Radoff: Awesome. And I'm Jon Radoff. I'm the other co-host of this program. I am the CEO of beamable, which is a game infrastructure company building the back end, the kind of the game engine for the back end, you could think of it. And I'm joined today by Jason Booth, who I worked with on, I got him to the right direction from my camera on Star Trek timelines this game right here. I worked on that for a few years together, but Jason has a long career doing a lot of cool stuff, even cooler than that. Jason, want you to give a brief intro of yourself?

Guest: I've been in the industry for about 32 years now. I started out doing MMOs, did all those music games guitar rock man, that kind of stuff. And done a lot of procedural stuff and generative stuff kind of along the way. Now I'm working for Meta. I also have 52 assets in the United Assets store that are mostly around generating and rendering trains and shaders and stuff like that.

Jon Radoff: And no shit everybody when I say that Jason is a world leading expert on this topic, I'm not like kidding, like he actually knows this shit. He's built the assets, he's built the tech, he's built inside the end, like he builds shader graphs, like he knows it all. So like, so you are going to learn something from this. I don't care how advanced a grab a programmer you are going to learn something from this today. So, get on this fucking live stream. This is we're keeping it real here. All right. In our audience by the word you're all game developers. It's fine if you want to see in about behind the scenes. That's cool too. Maybe you're a feature game developer, but this is really a great topic for anybody who is building a game. And last but not least, we got Oscar who's producing us. He's going to queue up the videos, the text. He's going to keep us on track. He's going to remind me of the schedule. He's going to go out into the comments. He's going to find out who's got a question. By the way, this is an open forum. So even though it's the four of us here on stage, if you post a comment that, hey, you've got something you want to say about any topic we're talking about today, could be game engines, could be procedural. So, in a way, we don't constrain the conversation. We might get into anything. Right. If you want to come in and join that conversation, Oscar will get you the Streamyard link. He can join on stage like it doesn't have to just be us. In fact, that's the kind of environment we really want to foster with this is more talking with you than talking at you. So thanks Oscar for helping me keeping us on track here. Jason, let's dive right in. I'll ask really annoying question. Unity or Unreal? What do I use for my game?

Guest: That depends on your game. The way I look at those engines, they've obviously changed over time. Unreal has very good tools, but you need to do things the unreal way. You need your little more constrained. It's much harder to customize an engine. Unity is infinitely customizable, but their built in stuff is not great. So if you're building something that's really custom, that you want to build all the tech yourself, I think for me personally, Unity is a more fun environment for that. But if you're more content focused and what you're building can be built out of the stock stuff, then Unreal has huge advantages. Unreal also is headed in a very specific direction around reducing content costs, through throwing a number of lights, any number of polygons at it. There's still restrictions, there's still cases or whatever, but they're really going down that route of you can just throw your content in and we will handle it. So that's a particular advantage with Unreal that really nobody else has right now.

Jon Radoff: And then so those are the two people talk about the most because they've been available commercially for the longest, but I mean there are these newer engines like Ado gets mentioned more and more. I'm just curious perspective on that or open source as a way to build game engines, which is kind of their strategy.

Guest: Yeah, when Unity was actively shooting themself in the face with the whole debacle last year, I went through and looked at a lot of different engines to see if any of my stuff would be, you know, warrant a port. I even looked at some lesser known ones like Flax, which actually if you're looking for Unity alternative, it feels very much like Unity, C sharp back end, you know, decent renderer, all that stuff. You know, Ado is a really weird one because I feel like it's very much like Blender in a way where the industry has this sort of standard ways of working and then this comes along completely different and feels very far and may or may not end up being a thing over time. I mean, there's definitely a lot of small games done in it, but I think right now it probably wouldn't scale to large productions. I mean, just some degree Unity is hard to scale to large productions as well. Unreal is the only sort of publicly known engine that I consider, you know, that scales well to very large stuff.

Jon Radoff: Let's talk about that when we talk about scales to large stuff. You mean big worlds or big teams both like big anything.

Guest: So big teams, because you have problems of data contention and how do we break up work, big worlds in the sense of streaming and stuff like that, like Unreal has some pretty nice streaming systems that they've been working on to allow all that stuff to work. It's also very battle tested on large scale titles, right, versus Unity, which is kind of a scrappier engine, but like as you ship in that engine, the wheels start to fall off of it and you have to like rebuild the car as it's moving, right? So, you know, so they and and part of that stems from the like the personality of the people building the engine and how what they were trying to solve when they first created it. I'm a big believer that the early code written on any system like that really shapes it long term. So Unity was trying to be scrappy, you know, have we just given up roll your own as a choice on the menu or roll your own should be a choice. I think people don't do it enough like IE my first 20 years in the industry that was the only choice. And, you know, if you're doing something that doesn't those engines fight against or don't suit well, then it's often a better choice. And it's a lot easier than people think, especially now when there are nice rendering libraries available and libraries for things that you can kind of plug together and build your own engine from.

Unknown: That was exactly where I was going.

Guest: And it feels like we're so too rich these days that, you know, that roll your own that seems like a better choice than ever. I have a confession to make despite having built a crap ton of games except for my own little personal side projects. I've never perfectly worked on a game that wasn't a roll your own engine like that's over, you know, a massive list of games. It's just like I've just always worked even small teams big teams. I've just managed to dodge. Does frostbite count as an independent? I mean, does that? Is frostbite count is roll your own? I don't know. I don't know.

Jon Radoff: I mean, if you started selling in the last couple of years, frostbite not may not even be something you've heard of before. Yeah, exactly.

Guest: Right. It's EA's own engine, right? It was funny though, when I first got to using unity, the most shocking thing for me is probably so mundane. It's that I could go new project and like make a new project because every studio I worked in, the first thing you did when you were starting on a new game was take the old game rip everything out of it. Like go delete a bunch of crap and get it all running again. And so like the idea that there was just a button that I could push and make a new product was like, oh my god, you know, so it's funny the commercialization of engines has actually pushed usability and things like that way higher on private engines as well. Like now it's not common or it's more common to be working with an internal engine and have the tools be nicer and like the sort of standard for those types of simple things be much higher.

Jon Radoff: We got a question from the audience here, which is kind of interesting. How do you incorporate Unreal with AI for AAA games? Hold that thought because I feel like we got to start from an even broader context first. Jason, let's come back to you in a little bit of the work or a lot of the work you've done on the unity assets store. So a lot of it has been procedural tech. Can you just kind of give us a lay of the land of some of the things you've worked on there and I'm curious if you consider that to be AI anything is that you may not have that is AI.

Guest: It's more procedural. I have always taken the approach of I'm trying to build better brushes because to me the interesting thing is helping an artist get out their intention. It's not and like how fast can you come out that intention. So like I made this it's an environment that are called microverse. It's entirely real time. There's never any baking like you move a mountain everything you know all the trees reappear like you move a river. It all just it's all in real time. And that's to basically keep the user who's working on building their game in flow right you want to be in flow state. You want to be able to move as fast as you can think about what the changes are. And so all of the proceduralism that stems out of that is from you know being able to define like how how this far is to look right and then being able to just like paint I want the far is here and like drag out you know an area and then have it just appear right. And then if you change anything it all changes and that's for iteration speed and the idea that you can iterate up to the very last minute you ship your game in a non destructive way and never you know because like if you've worked with a lot of train engines if you if you want to move where something is it requires like first for pairing where it used to be right and getting it back to how it was and then going and moving it and it's very time consuming. Which whatever tools you're using when that's AI procedural whatever it's really about getting somebody's idea out as they see it because the interesting part to me is that human part of this is what I have in my head and I need to get out right it's it's why like you know when when AI first came out never does text prompting stuff it's neat to get something right but it kind of it doesn't last that long as an entertainment because you're kind of getting it back to you. Because you're kind of getting something that sort of you don't feel like you created it right and you don't have that personal connection with it and I think to some degree the audience has a harder time getting a personal connection if the artist doesn't have that personal connection and so yeah so that's why favorite proceduralism I do think there's an opportunity to combine those because AI is very good when it is allowed to make mistakes and proceduralism can enforce rules very well right. And so the idea of letting AI be sort of a high level construct like somebody you know AI generating the idea of a D&D campaign but then the procedural rules turning that into the 3D graphics as opposed to just having you know one system try to do everything is kind of an exciting intersection of those texts that I think will will be interesting to see.

Jon Radoff: Even if even the free intersection between those two though I guess one thing worth commenting on is we're in this era of now everyone's an AI company that's hot so right you see all this proceduralism out there and suddenly it's AI. It's categorized as AI.

Guest: I mean we're sticking we're sticking AI on anything that's you know it's 500 miles from AI. It's the best right it's you know it's eyeball aggregators in the 90s it's just the latest buzz thing and so everybody's trying to classify their stuff as that because it helps raise money and you know make some look cutting edge but you know but maybe it's a little more than that because I mean we legit I can go reach out you know my AI pearls of wisdom book over here I mean we used to call procedural stuff AI and games for a long time like you know I mean that's the light was the way I know we're going to do it. The AI now means you know LLM right right and that's not which is the AI means LLM now.

Jon Radoff: Paul when when does candy crush just run an AI that figures out how to extract money I know you can't tell us about like but you said one interesting thing earlier by the way which is I didn't realize that you guys were so oriented around sort of homegrown engines for even a game like that. I think I heard that but then you've you've made use of a combination of proceduralism and I feel like the comment Jason said about like giving the artists a better brush knowing you I thought that really resonated.

Unknown: Oh I did I felt like I hit me right in the fields like like going back to the very genesis of my career I think so I come from a you know a very early enough in the process where we were we were not trying to make games this content right and so we were trying to make the value prospect.

Guest: The player was a little different we were trying to make games that for your 49 95 like where is big as possible right. And so it wasn't just the I mean Jason's examples what lay in the art domain but like over in the cut over in the game design coming we were trying to build I mean random mat generation was everything for games that you know I was not an optional feature right you know and so my early part of my design career was like how do you make the game. Infinitely replayable like you know rogue likes are my you know my my ear form of game design because you need to build a procedural system that that that generates an exploratory space that hopefully goes on for you know goes on forever right and I've taken that ethos you know I'm built a career around it largely I mean these days we some of that procedure work. And so I think it's hidden underneath the engine you know less I mean I don't think games are games are not infinite like in the way they used to be except you know in the dark like the games I'm working on are not infinite that way they're usually infinite by my virtue was putting content into them but on the backside we build I still build with an infinite game mindset because the more I can procedurally decide how something will get built the more effective it is right you know I think a surprising number of the games you played.

Unknown: It's really easy to imagine that you're you're cutting edge mobile games or you know or your big quadruple A games are oh yeah they decide at once and they sort of just pump content into it is so much more messy than that like we're over their hammering on Royals we're not it's not an automotive assembly line like you think it is it's role the spoke roles Royce is being hammered in the shape far more than we'd like to admit right and I think I proceed that procedural thinking.

Guest: I think there's another thing to which is the procedural you don't see so like for instance all of the rock band animations of the characters playing their their instruments is all done from it's procedurally generated from the expert tracks of the game so when those gems are coming down for your drum or your guitar part that's actually driving all the animations and the bends and everything because basically like we didn't want to have a content pipeline where we had to motion capture or like hand animate all that stuff right so it's being used in a way that doesn't look like you know it's not world creation it's not level creation or whatever it's just like we want a fast content pipeline and because of that rock band has 5,000 downloadable songs and guitar hero has 300 because they had the motion capture everything right so like those choices about you know which techniques you use to build your game. You're you know result have massive impacts on what you're able to do if you have a success. So from my side I couldn't agree with you more Jason you're speaking my language like like like so from my perspective I don't get to do that nearly as often as I like even though I think I'm pretty good at it because in at least in my experience the engineers and the designers sit too far apart and don't even think this way that's my experience I don't know man how do you. How do you think about it. I am a big believer that having design separate from everything else is actually a problem and one of the things that was really interesting in the early days of harmonics is that we didn't have designers because everyone was in charge of the design. There comes a point with certain types of games where you need somebody who's going to go and you know build quests or you know you need specialization but I think the thing is to really have those teams work together on solving problems not that sort of throw it over the wall like oh I'm a designer and here's what we want to do so you know you go build it because often engineers see tons of low hanging fruit and designers don't know how hard something is and so they make something that's actually really hard when something else you find. And so like you really have to you know and whether that's you know synergized in one person or across the team it doesn't really matter but you really have to have everyone involved in the design and understanding both the what matters about the goals but also what's possible and what's easy and what's hard to get the best result so. Yeah it's not just it's not just like oh they're not sitting together it's like no no no you guys need to live like in your pod together and hear everything that's going on like I never wanted an office because I like to be in the middle of the art pit and here what's going on you know. Yeah I call it full stack design.

Jon Radoff: Yeah well let's go back to David's original question that went into this I mean they even broader than unreal unity whatever the major engines are. I mean I'm not even sure what we mean by AI in this context because it could mean anything from a path finding algorithm for a character to what people are currently talking about AI which is like making a 3d model from a generative prompt or something or anything in between those but just I guess Jason the question more of a lay of the land question like are you seeing interesting stuff with AI how or proceduralism even however you'd like to classify it in any of the major engines.

Guest: I mean proceduralism is everywhere you know unity's PCG graph is very similar to microverse in a lot of ways it's just a matter of like how do I manage this content and produce things at a scale that I need right. I do think like you know the obvious one with LLM's is like oh you know NBC's that can interact on a better level or whatever but I think we're going to see these coming in and I in all kinds of different cases but I think the success is going to be how much of a brush is it as opposed to how much of a generator right like how specific can I get how much can I edit what I have to get the actual result a lot of. Experimentations I've seen in AI have the ones that have failed is because they have very specific requirements and the tools don't let them really get that specific and the ones that have worked has been many either you know okay we don't really care about style so much so just generate or we custom trained on our art style to produce variations on the theme that we spent forever getting right. You just pointed at the thing that I think is that the heart of maybe the thing I'm most exercise about the moment I want to take us I think we've all were there at the time want to take us all back to desktop publishing came out and we all got the ability to make just ugly ass documents. Like like everybody became a publisher and you remember about five years everything looked like crap right you know and when we sort of democratize the tools what doesn't get democratized is taste right and I think we're kind of we're entering where we can see there from here with with a lot of this AI gym material like I've already gone through the uncanny value of people I think your AI gym in your PowerPoint deck is more of a signal that you suck at prompting than it is. Hey I'm smart about using AI at least in the next time seeing right and I don't know I think the bar for taste is is it as important or more important that it ever was well there there is a danger there so like I do a lot of work with music I'm from music school and stuff and so what you're seeing now is that many of the jobs that could have been done to keep you afloat while you work on your career are going to be able to be done by AI you're not going to produce music anymore. You're not going to you know fillers filler music for backgrounds like you can just generate all that stuff now so when those got jobs go away and and imagine that they go away for art as well how do you get experts in taste and design when they don't have a pathway where they can earn a living while they build that right and so we may end up in a crisis of of taste eventually essentially where we don't have you know enough people that really know what they're doing. So how to get the look correct right and so then you get a deduction in quality which again you know it's like people listen to MP3s instead of waves right there's like a level of quality we're willing to reduce for quantity but when I look at the overall content landscape in any industry quantity is not been our problem quality problem. So coming up with ways that we can generate more that's not at a high quality level doesn't really get us much. And it enables a few at scale game types that we're all we've all sort of families about but like you're gone with you 100% like that's actually where I was going with that is that like it feels like we've always had enough people to make quantity with the exception of building the holodeck right you know and you know of course we're all excited about our holodecks but I don't know and I'm.

Jon Radoff: I see that the crisis of taste coming and I think it's not going to be a thing to start you're right it's it's why to me the interesting thing about AI is how can this be a brush not how can this just generate what I want as an result right I love the analogy got to do a quick station identification here and then we're going to introduce our first live guest who communicated with Oscar just like I said and is now going to come up to the stage but let me just comment on that that's a good reminder. That that's a good reminder that this is your program you can join this you can be part of it live you don't have to. Pre-spec scheduled as a guest Jason was pre scheduled as a guest but that could be any of you particularly during the live program so if you have a topic you want to bring up that's contextually related to anything we're covering by all means like join us turn your camera on and you can be part of this broadcast. And hey if you have something that's a little bit of a different direction you want to talk about you think it's super current for this week. Come on board like we want to hear from you this is this is your platform but let's introduce William Volk here William thanks for deciding join us. Super quick intro to yourself so everybody knows where you're coming from and then what do you have to add to the conversation.

Guest: Well I'm going to walk I've been in the game industry since I was in grad school so I've been in the game industry probably as long as anyone has first game with that. That's a tape your system in the background yeah that's my that's my office by the way just kidding. I click shit no no it's not it's not a big silly but yeah and I you know I started out at Avilan Hill as a play tester and then I built free games in them and eventually I me and her done to act a vision got very lucky because not only to get hard to act a vision but I knew Bobby Codic before he took over act a vision so we had a product that had failed totally Bobby kept me on with a really successful product that took over a hundred CD's for the PlayStation that were sent to schools that was light span then I went into mobile then I did some work with the Shibuya people. You know pretty early a mobile as well I did mobile games with the Wayne's brothers before the iPhone ever shipped currently working on a real money casino as well as my casual game stuff which is a movie trivia game.

Jon Radoff: That is how much to you don't know Jack in the way with all right so what and you've been here from the beginning and you're going to help us shut down the industry with with web three of snow and I have a question no no no no it's your thing I was on a conversation this morning with someone who is a principal at a two point something

Guest: like a $1.2 dollar casino company and we were just laughing because we were talking about how early crappy the web free games are this was the conversation it's like oh yeah yeah that only people reason the people are playing that game in that country because they think they can make money and I've had this discussion over again.

Jon Radoff: I don't know how to do that that's that's already a business so you might as well just you know lean all the way in so William what did you want to bring the conversation.

Unknown: Oh yeah I actually work on the game's yeah I worked on procedurally generated games early in my career even though Wikipedia disagrees you know what happened is it was like what else could I do you know let's take like for example the original Mac 128 kilobytes of RAM I did a product called the perimeter of peril and I couldn't possibly you know have the levels exist.

Guest: Before I did the game so I accurately generate the entire levels all the passages doors objects creatures and so on completely random eyes because it's an overwave doing it that the whole story behind that game but yeah so I want to say about a I couple things one because of deep seek the venture community is sort of their opinion is shifting on a I a little bit they're not as gun holes they're not going to be a little bit. As gun holes they were a couple weeks ago it's really funny. That's right.

Jon Radoff: And they're willing to like that to me the question there is why because the cool thing about deep seek is the ability to run that model on your own PC and pretty much a very mediocre PC can.

Guest: Right it's a very low model god if you can only do that with bitcoins right but it's like now you've got that you've got actual ability to run.

Jon Radoff: For angle models on your own machine you could all true which gives you the ability of integrating that with a game here's a fact of life that's not a way to build a business.

Guest: The way venture capital works is go to a soccer game with six year olds and wherever the ball is all the kids are going to run to that ball they don't hold their positions right so venture community was super hot and AI and when Nvidia's tanked in. So we got you to be all that sort of went down they kind of nervous it's so weird I'm sorry this is what I'm hearing you. But yeah what's what you just said what you just said you know up until deep seek they thought there were 15 horses on the field and they just found out there 500 horses on the field and they got a needed moment to regroup and come back. So the valuations you know suddenly it doesn't take billions of dollars to do it I mean they they poured they poured a lot of money into companies that said they need that money to build what they're building in deep seek sort of blew that apart just an observation.

Unknown: I will say this much show about a year and a half ago I was on a bike ride like I usually am and I was having coffee and I ran to a guy wearing a t-shirt from you know a rock star rock star games and he told me he was retired and all that and the question I asked him I go.

Guest: How far away we from using LLMS for the characters because if you look at like sort of like the cost of building the games and all that one the big things now is getting all that dialogue in there and programming all the dialogue. Can that be done well you know it was a difficult problem before deep seek now it looks like you could pull it off you could actually ship a game with some level of character intelligence and conversations that are basically I mean people are I don't know I don't want to mention names one person. It's very anti-AI in the gaming industry is like saying oh my god it's going to be horrible they'll they'll say terrible things that's help people to kill themselves and blah blah blah blah you know yeah that's possible.

Jon Radoff: I'm not so concerned about the saying terrible thing like okay you probably that probably is a minor issue but that can become a big issue of the exact wrong thing I said but to me the opportunity with LLMS isn't just make my boulders gate for boulders gate three my favorite game of all time personally. When there's gate for some day which I don't know that that will ever happen but I'll dream about it happening but I don't dream of it happening with all those characters being LLM driven I think first of all there is something to be to the very crafted experience of those interactions to the gameplay experience is not about maximizing my my NPC dialogue time with those characters it's about getting to particular plot points so I don't want more conversations I want meaningful conversations or the NPCs to act in meaningful ways. And yeah what and three there's sort of something to be said for the shared experience to the fact that I can talk to my friends about the same thing that we did. That's the point we made I think are important from game is I do think LLMS could be really really interesting. I'm just going to pull new categories of games that aren't really possible right now without that but it's not to make my existing NPC games like ten times as long conversations with with the characters so that we can talk about anything whether it's relevant to the plot or not. Oh by the way remember when I said we wouldn't talk about AI on every episode of game LLM live stream we've now we've now hit that prediction 0% of the time so.

Guest: I came in here and that's what the thing I miss yesterday is it would be much more fun to talk about web free and games and this and this some of the web free games around. That would have been more enjoyable.

Jon Radoff: This is not the web free.

Guest: No no no we're done with this. I just just came up this morning and I had to bring it up yeah. I use LLMS mainly for things like I need to generate meaningful movie quotes from 150 movies do do give me an Excel file that does that and I have to check it but it does save a lot of time. To do that like or what what's a good short description of this movie the one interesting thing that hit me in AI recently is a thing called I'm going to give them a plug because it's really fantastic it's called any male finder. I don't know how they pulled it off but what they're doing is they're using AI and they're generating emails for people you want to get in touch with that are verifiable that when they say it's a real email it's a real email that that person and. I think they're actually hitting their the male servers and that's been around for a while I wouldn't really well rocket reaches a piece of crap compared to this so just want to bring it up yeah so they claim you're doing something in the AI all I know is it's the first time I've ever used one of these things that actually worked because most like utter garbage yeah so if you work at at screen rant or rotten tomatoes please don't be mad at me because I probably made. I probably made you today everyone in the involve with anything there because I did that today to you sorry I don't know let's piss off some some different companies Jason what's what's wrong with game engines today. I mean it depends on what you're talking about you've all got problems you know and there's a lot of it's just general software problem like the bigger the thing you make the more you know problems you solve just the more bloat and more complex. You know and and we're also just seeing engine teams like you know an engine team for like in harmonics we were like 10 people and then you come along and hits 8,000 people how do you you know how do you even do that right like no wonder they're having problems and the range of platforms keeps getting wider which makes the problems harder you know so we're going to get a lot of things that we can do. It's a lot of problems harder you know so I don't know if there's like a what's wrong with game engines it's just like what's what's the challenge set with this particular one you know.

Jon Radoff: I have a general thesis of software which is it always mostly gets worse it has better features that it gets over time but software in general it seems like it gets worse the longer it's around.

Guest: It's more complex with every feature you add it yeah it definitely gets worse there are definitely moments in engineering where I've had where like I suddenly realized that the thing that we've been doing forever could all be done in like a page of code right and I've had that happen with like the camera system that I built for disruptor being was based off the one I built for harmonics which is based off of years of writing camera systems and realizing that like okay this is. All this really needs to be but a lot of times you have to write a thing a bunch of times before you get there and so like these engines have these big systems that have been written and there's it's very hard to change them it's very hard to not support a feature you used to so they're kind of locked into this accelerated collection of craft.

Jon Radoff: Which engine company do you think is best position just in their DNA to take advantage of new AI technologies that we've been talking about.

Guest: I don't know you know epic is is kind of gone another way in that I'm sure they're exploring AI and stuff but their whole thing has been scan the world. Proceduralism make anything run in our engine no matter what so I think they're less interested maybe in in in like a content pipeline built out of generative AI. Because of the problems it has versus what they're doing in the momentum they already have. Unities you know going along that route I actually think the more interesting things are going to come from small applications that do one thing well then they are from the big a game engine companies.

Jon Radoff: There was interesting report. I think it must be a year maybe two years ago now from a 16 Z the big venture firm they've got a gaming fund maybe you can locate the report Oscar and bring it up but they did a report on game engines. And there are general take was that I mean of course they're investing in the next new thing so they're kind of in the business of slamming the incumbents. Sure not necessarily super close to throwing the incumbents out in my opinion even the ones that fucked up big time with business models like one of the two big ones but. But they had kind of a thesis that the game engine of the future almost has to be rethought from the ground up for a couple of reasons one is procedural and content and generative content creation is going to be so inherent to the creative process that you're constantly iterating around that in the future was sort of one aspect and the second is that the game engines as they are now. From a from a developer standpoint are not really multi user multiplayer products like they envisioned that yeah are in the engine together working collaborative generating things putting things together in real time in the engine with a team is supposed to that whole complexity of team dynamics that we've talked on in first couple minutes.

Guest: Well we do have a rise of a different type of engine which is basically spun out of dreams on the PlayStation. I don't know if you guys ever remember that game but everything is very built on a very different renderer now we're seeing things like unbound which is a sign distance field you know modeling construction set. And those are like dreams honestly kind of hit the metaverse goals that everybody is shooting for early and kind of wasn't a success because of a bunch of reasons it was kind of too early. Yeah metaverse like a zero billion dollar business everyone thinks it's going to be great but it's like oh I totally disagree with that like arguably the biggest game engine is actually just Minecraft. Yeah yeah my craft is great my craft and also Roblox Roblox is a big game engine.

Jon Radoff: Minecraft Minecraft is the metaverse let's just right I like that you don't have a better metaverse. That's right.

Guest: Yeah well I think interesting because minecraft. The commercial metaverse sorry Jason. I mean there's a bunch of companies trying this kind of thing but like the thing that made minecraft work is that your core primitive of a block. Is easy to understand easy to rock we saw all those minecraft minecraft clones that tried to do like smooth voxel stuff and they all failed because it wasn't the representation being pretty that mattered it was the. The feeling of oh I created this thing and I understand how to create it almost immediately because the only thing I have to understand is a block. It's an example of where constraints can be a good thing the fact that it's all blocks actually is constrained it turns out it allows people to not have to worry about a lot of things they can just go ahead and build stuff immediately it's easy. Yeah and and yeah I was going to mention that there are some small game engines that are specific to a certain genre. For example for visual novels is Ren pie and Ren pie is fine great I used it for a product it's fantastic and what it does you know because it's so focused at that one thing right. And people like that turnkey stuff you know but that's why I was kind of alluring to earlier that I think like if we end up with an engine that's like like I don't think the big engines necessarily will be the ones that come out with the interesting AI stuff it'll be more some small team on a very specific vector right that they push that thing as far as it can go in their own language and their own stuff as opposed to like oh we're replacing traditional systems with the. Like AI enhanced ones so what happens when the ultimate engine is no engine right I mean that's where some of the I mean that's the tacked on the big future you know the bigger rights and right now like you know what happens when I have 60 frame a second mid journey right you know that's using you know player inputs as one of the context for the image and we're already seeing test cases of this like the what was the do not do Marl but the the basically the. The imagined doom game that doesn't actually have a game in jibah and it's just the image generation LL that's been trained on the game do right so when we have high frame rate high frame rate image generation with player input as the context don't we have. I still think on a silly notes I'm going to do them to run in a PDF yeah yeah I still think that's not a pregnancy test like it's crazy. I think that might not be as interesting as we think it is like okay why so like if there's a guy I can't remember his name right now but he posts on Twitter and lose guy and he does stuff where he'll model stuff really quickly but then he doesn't usually in dreams but he also uses unbound and stuff but then he'll have AI enhance that right and I think I still come back to the premise that I'm skipping. I'm skipping the word right now my head for it but like ownership and you know building something that matters there's not a potentiality to it is that we're going to get that matters and that reads through a piece of art as the interesting part even if you can get the quality to the same quality bar right like even if every frame looks just as good as reality that an artist intention is going to actually be what makes. I mean you see that there are a lot of games that have a very stylized look that doesn't look anything like reality to really well you know sure yeah yeah yeah. I guess I guess I'd ask Colin Porton as a director to a movie like I think you can project artistic intentionality onto a thing without a different layers of the experience like it. Yeah so but like we you know you look at golem right golem is entirely CGI why does it work because you have any circus in there doing the acting and bringing him to life and Peter Jackson directing it yeah yeah that a good script and you know all the bits but like you know they chose to have somebody in a suit no no. I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I think I

Jon Radoff: I think I think I think I like having when that happens I see the Lisa Chen suggested she's trying to do something beyond the τηνning into ideal values she's trying to do something beyond the

Unknown: でした of general values ​​of the digital http

Guest: lot of information towards the work that is trying to get through all of the researcher just an expert called Linda a quality bar and everything. But there's something about it. It was a performance capture that was then remapped using like, I think original bus that they had of what's his name, the grand admiral, the Eddie, you know, from road one. Yeah. Yeah. Yeah.

Jon Radoff: I do feel like there is something to this idea of that A16Z is talking about, which is sort of like the multi-userness of the of the building environment becoming important to the crafting process. And I feel like that's where Minecraft actually there's the primitives back to the building with simple blocks and the fact that like the simplicity of it made it so much more accessible to people. Like that's a huge part of it. But then that you can just jump in with your friends. Like my son is a huge Minecraft player. He's like in this game. I don't know why you can think thousands of hours at this point, but he but he's in there with his friends and they build stuff together even when it isn't like go and code something and Java to add a feature. It's just building and collaborating together. And that's an aspect that's missing from the commercial engines that I think is interesting to think about. And that's where the AI and the procedural stuff you bolted onto that. It could you get interesting. Paul seems like you wanted to jump in there.

Guest: Yeah. I wanted to throw in the neutral maybe because like I think we're back to special case, right? You know, a lot of the games, you know, it's really easy to point to Minecraft and I agree with you by the way, Minecraft, you know, is the metaverse or Roblox is the metaverse. I like, you know, those are those are provocative and pretty true thoughts. But like you also invoke Baldur's Gate, which is famously a single player experience, right? You know, the game I work on is a famously mostly single player experience. Like I can go down a list of like most profitable, biggest, you know, use whatever criteria you want, you know, experiences and the social games are super present, but they are far from exhausted, right? You know, for and so I'm not sure, I mean, I agree with you for the specific application. I mean, I was working on a game like, you know, those kind of games work multiplayer. It's a social context. Yes, but I don't know why I need what you just described for to build Baldur's Gate 4. Yeah, there's maybe there's been a number of engines that had like kind of multi developer things and it never took off because the reality is if you're doing pro work, you don't actually need that most of the time. But like if you're doing consumer level stuff where it's like, I would get together with my friends and social experience,

Unknown: that's totally different. Right? Yeah. Totally. I did reason like Google Docs lets you work together, but Photoshop doesn't. Right? I'm just imagining Photoshop and Google Docs trying to

Guest: somebody else, stop it, you know, like, you know, living nightmare. Yeah.

Jon Radoff: It's interesting though to then look at something like UAFN where they've tried to lock it, lock down the environment a lot. Like I like the idea of UAFN, the ability to like build around this core of gameplay that obviously 100 million plus people love doing all the time, but like, like every little thing every asset you make has to go through an approval process, which is like the next level, it's like the next level from the App Store. Like first we had to fight with Apple

Unknown: to let them just release the damn game and they'd comment about a couple of things they run into.

Jon Radoff: Now it's like every asset in the game has to have an approval process and that seems like, I mean, I have two two kind of thoughts on this. One is that's just an annoying process. So,

Guest: I mean, is it all about time to penis or, you know, yeah, exactly. Probably.

Unknown: Yeah. Probably part of it. I mean, it's also an impediment to like how could you ever do

Jon Radoff: generative or even procedural content in that in that environment if you've got to put all

Guest: these generations into a approval process? Maybe precision. It's not a edit time operation, right? Like where I'm using it to generate a specific thing and then the specific thing is what ships. You know, runtime generation where it's like, okay, well, you're going to produce variance on a theme or you're going to produce whatever the user wants. That's a whole other thing.

Jon Radoff: I don't know, you know, it's like, what's the problematic actually because runtime generation, I'm not sure actually their thoughts on that, but maybe it could be constrained if they see it as very the healthy even QA that like, but also like what will happen is the amount of content entering

Unknown: the approval process will explode by some huge percentage if you have a supportive content.

Jon Radoff: It's like, they've approved already 1000 mid-journey images. Yeah, well, we've already seen

Guest: the robots. The robots approving robots all the way down. Yo, I got hurt like robots. You know the assets or and Epic sort of been flooded with AI generated icons and stuff because basically as soon as somebody was like, hey, I could just hook this thing up and generate packages

Jon Radoff: and ship them. Somebody did it, you know. And if it's a web 3 game, it's robots playing your game too. So, we've got so many bots out there. I actually know some people building web 3 games that are quite genuine, but that is a typical use case for many web 3 games out there.

Guest: Certainly for a lot of the web 3 metaverses, a lot of them are bot city. Well, it's because we're harvesting activity and not engagement. I mean, I think we're, you know, we need some currency of authenticity. I mean, I think we're headed toward the authenticity. I've already mentioned it. I think we're headed toward authenticity or a crisis of style. I mean, you think the value of authenticity rises in this environment. But I really, I guess, if I want to shout out anything from this conversation before it's over, it's a love Jason's idea that like what happens to the when the middle and gets pulled out of like, how do we map? How do people achieve a level of that when the intervening steps are yanked out of the middle? Like that's a pretty provocative. I do think the business model is at play here. When we made games that we're about producing the most game, the most fun game we could, there nobody was botting them. As soon as we put long curves like early MMOs with grind or, you know, long curves for monetization reasons, etc. than everyone bots them. So, I'm, I am not even though I've worked on all those models and stuff. I don't love them because they are making us produce games that are more compulsion than they are joy. And ultimately, like, I got into this industry to make the most fun things possible and to watch people light up from that. And I think like as long as we're using free-to-play models, advertising-based models, you know, any of these business models that we're doing now, you're going to see botting because

Unknown: people are just trying to get to the fun part. True.

Jon Radoff: True. So you can go ahead, John. Go, Paul. I'll just say, well, the botting is only bad when it's bad.

Guest: And what I mean by that is not every game model actually requires a level playing field competition. Like, bot, bot suck when you're trying to compare your work to another person. They largely are relevant if I'm playing a solv-back to Baldur's game. Why do I care if John bots his Baldur's game three game? I guess we're back to, you know, game form, game form matters when we start talking about a lot of these concepts, right? We always forget about solitary games. And yet, I feel like a lot of people- Yeah, yeah, I mean, I made the mistake once of taking the example of the share drawing game that was really popular, I guess, in 2000. And launching a work game that was too play, too, too player-only and stupid dumb mistake. A really brilliant sort of a work game that was like, you know, like the ones where you try to find words, but it was basically balloons. And when you find a word, they pop and new balloons would float in. It was fast. It was exciting. And we screwed up totally because we took the example of draw of this as like, we have to do

Unknown: that. But no, we didn't have to do that at all. And then another example, we had the best selling

Guest: botchi ball game on the iPhone. And we wasted a lot of effort making it multiplayer. It was, it's a botchi's very tough game. That is a very specific deep cut, William. Well, the botchi game, for example, if you're playing botchi ball, the actual botchi ball game itself is a difficult game to do multiplayer because your turn is very short or it isn't. And what we probably needed to do was actually create a botchi-like game for the multiplayer aspect of it and

Unknown: keep the regular botchi ball game as a solo game. But you know, that was a million download game

Guest: with no advertising. And we did so many stupid, well, there was a stupid, there's a whole bunch of stupidity. You forgot to ask your, the million people who care got it and then you forgot to ask them what they were doing. Well, here's the dumbest thing I've ever existed of mobile games, maybe the dumbest. It hit number one, the Italy without even being mentioned. And no one offered, we were so tied to a license we had that got the company sold. So I sold this company based on a license I had gotten for a particular poker game. And because of that, we were so focused on that, no one had the energy or the effort to take the game and convert it to Italian to, to localize it, even though it was number one game in early for some time, you know, it was like, like, dust. I could do a, you know, yeah, the, the, let me just go into a GDC, Rand, GDC doesn't like to hear stories about failures. Otherwise, I would do that, you know. But there's a good use case for, for your AI plugin. Oh, yeah, AIZ makes it, makes that stuff cake, cake. I was speaking of the off story, I will say the current mobile game, I, the game I have, which is Web and Mobile, is actually created, the original was created in 2015 because of an Apple rejection. So Apple rejected a very funny, humorous game. And we took the same engine and turned it into the movie trivia game because Apple can be difficult. So now when I build stuff for Casual, I'm doing Web, it's slash mobile because I never can tell what Apple will, or Google will do. You can't, you can't

Jon Radoff: determine that. I'm aware of a project that was working on a James Bond game for the app. But needed to go on Apple Arcade. And the game was basically, I don't know, a few million dollars into production. And then they brought a tap on Apple's like, can you take the guns out

Guest: with this game? Well, that was the old, that was the old, the old Mortal Kombat. The old Mortal Kombat versus Sega versus Nintendo thing, if you remember. That's what made, that's one of the things that made Genesis big was the fact that Mortal Kombat and the Genesis had blood. With the cheat code. Yeah.

Jon Radoff: Right, Jason. So I'm a good story, by the way, that James Bond story.

Unknown: For a student who's maybe, I don't know, they're in college, they're post-college. They want to,

Jon Radoff: they want to make money as a game developer. They don't want to just be someone messing around with an engine in their free time. They want to work in the industry in a game. What should they learn and engine? Where should they go? They should learn as much as they can and produce actionable

Guest: like results from it. You know, I've, I did some lecturing across a bunch of schools for a while. And I saw such a range. I saw one person come up to me who told me they wanted to be a lead designer. And I said, great, show me something you've designed. And they're like, well, I'm not a programmer. And, and I'm like, what year are you in? He's like fourth year. And I'm like, you're never going to get this industry because like you haven't taken the first step. So ultimately, what, what I look for when I see young people is like this question of, are they going to do something anyway? Are they going to do it because they just love it so much? And the people that I've hired who've been like that have always worked out because they're just going to work at it and figure it out. So, you know, I don't think you can build your own thing. You can learn unity, you can learn unreal, you can learn some obscure engine. But at the end of the day, what you want to be able to do is show people cool things that you've built that show that you work through the problems. And we'll figure it out because the challenges that you're going to face on the next game have nothing to do with the challenges you face before. It's like a shipping is your best feature. Yeah. And it's real geniuseship. Right. I like that. And I look at it like a stand-up comedian who like builds up this routine, does his Netflix special and then throws out every joke he's written and starts from nothing. Right. It's like because you know, it's not like I have some special formula because I worked on guitar hero or rock band that's going to make the next thing a hit. It's just that I'm going to work my ass off to do the best thing I can do. And I'm going to do that every time. And so like those are the people I'm looking for. And if your results show that, that'll get you hired. It's your library of processes though that you're carrying from project to project. Yeah. Your process is your expectations. Patterns of tech or whatever. You get better at it. But like, you know, there's no guarantees in this industry. There's too much too many things out of your control. And like anybody who thinks that, you know, I don't know, any particular higher, whatever is going to give them some guarantee. It's like that there's a line to themselves. And people love to lie themselves. But the people that are just like to build things anyway, they're always going to build something cool. Whether it's cells or not, that's another thing.

Jon Radoff: I love the shipping comment. It's the it's the biggest thing I tell to everybody. So

Unknown: I have a thesis that just the core of successful teams is just constant shipping and your

Jon Radoff: velocity in the development process. If you can master velocity and ship, you can even, you can just start with crap and you'll eventually get there, right? Like you can get there. Most of what I worked on was crap for most of the way or even 100% of the way. And then we even fixed it later. But, you know, I feel like that's just how game development works a lot of the time.

Unknown: But where it doesn't work at all is it's a GDD or a pitch deck and you're constantly iterating

Jon Radoff: the pitch deck in response to what other people are telling you about the GDD and the pitch deck. Yeah, just like the pitch deck I get, you need one to like get the attention of a publisher or a VC. But what you really need is playable product. Because I feel like what impresses the investor, the publisher, more actually is they saw the pitch deck with the build at the time and then you go back to them a week later and you're like, here's the new stuff that we just shipped. And then it's two weeks later and we just did another one. And here's what we learned and like constantly ship and be learning from the building and the costing process. I also feel like you can't just go in with

Guest: the pitch deck anymore. No, no, you can't. It's very least you need to over. If you have a video that

Unknown: portrays to play it again, that might be enough. But you can't go in with just a pitch deck. I think

Guest: I think it's a real software or go home. Like I feel like, you know, people are, people are interested in it. I mean, all those people we're talking about are those, is that fourth year student that Jason talked about?

Jon Radoff: They're all that. Gary Glen Ross version. I got to write that's about shipping the damn game.

Guest: Like, did any of you guys remember Prince of Persia's sands of time? Of course. Yeah. Obviously, fantastic game. The first thing done of that game was an animator animated the prince running, going up the wall and jumping to a platform and pulling himself up. And he happened to be roommates with one of the programmers at the company. They were kind of an untested team. And they built that. And they were like, this is what we're building. That's right. We don't need to build something large. It can be hacked together. Like lame ships. That was a big expression in harmonics. Like nobody's going to see how this is made. Show the experience. Get it in some way that somebody can understand what you're trying to build and why it's special as fast as possible.

Unknown: Hey, I mean, I think we worked on about like five versions of this game before we built the one that actually was going to be any good. And I know you were not. You had issues even with what we

Jon Radoff: did ship. But we had. Yeah. But we like we did. I mean, the only regret I had is like, I wish the five versions that were completely not the right idea. We had figured out a little bit faster. But like, but, but we that's kind of the process of game building. But you find that out from shipping. Like we could have made the most perfect pitch deck and an amazing GDD about any one of those five, even though it wasn't really playable when you got down to the experience of the game or just wasn't good enough. So just ship. And speaking of which, we're at the top of the hour. So we've shipped episode number three. And we're going to keep shipping. We're going to we're going to go with exactly what we just proposed here. We're going to ship every week. Good or bad. I think we're

Guest: doing pretty well. We're bringing in. This was enjoyable. Yeah. Yeah. So join us again next week.

Jon Radoff: It's going to be the same time one PM Eastern on Thursdays. We're going to do this every single week. And we're going to build the biggest audience out there of game developers who are building games. And I hope you found this really useful. I think there are some real nuggets of knowledge here. Jason, you're welcome to come back anytime you want to talk to us and going in the conversation with game developers here. You know so much having worked here, you know, for a while and been deep in in the tech, but also the craft. Like I think in the intro, we didn't pay enough service to not only the deep tech you know, but the craft of game making that

Unknown: you're so deeply involved in for years. So thanks for joining us. Yeah. Thanks for having me.

Guest: Thanks for having me. I really appreciate it. I will just add that my biggest success in games came after one of the biggest failures. So I agree with the ship ship ship and learn from it. Yeah. Yeah.

Jon Radoff: You haven't had a big failure in games. You just haven't been out it long enough. Never got it. It's just a focus too. What can I say? LG LP too. That's really sucked. Yeah. Yeah. Thanks. Thanks everybody. We'll do it again. Thanks for being on. All right. All right. Bye. God. Have a great day, everybody.

Unknown: Thank you, John.