Creative Future with Artificial Intelligence

[Music] thank you good morning everyone thank you for making it to day five of GDC you've survived this long I appreciate you being here despite the the hangovers the sleeplessness all of the exhaustion that you've endured hopefully from making Amazing Friends good memories and uh you know amazing Creative Connections that are going to last you way longer than you know this week um a few housekeeping things as they usually ask you to do obviously silence your thingies uh fill out the form but the one thing I'll actually ask on top of that is we're going to attempt to do some live demo things here using the Wi-Fi from the conference so if you're using that and you could turn it off just for the duration of this it might bode well for for this presentation uh yeah but other than that let's uh let's Dive Right In so as you can

hopefully tell from the slide this is a sequel which I know we all hate sequels at this point but uh there you go we've made it all the way back to GDC as well to a talk I did back in 2019. and then that was kind of a sequel to a talk I did I did back in 2017 and then that was kind of a sequel to a talk I did back in 2013. so it's interesting because looking at all of these years standing up here and talking about creativity and technology and their relationship and how it can serve Us best and how we can see through a lens that is most impactful you know I can't help but feel at times uh pretty much like this meme uh and you know it it makes sense at the same time you know part of me is a little bit offended but I did not fix the industry in like 120 slides but uh you know what are you gonna do so we're gonna try again but uh but I think about it a lot right like the talk from 2019 went through all of

the cool AI things that you can expect and what they're going to look like and it was back in the time when we weren't talking about it now you can't you know walk past the toaster without it telling you about its new AI feature so it's just everywhere and uh and then my train of thought has to shift right like how do I be most useful to you why can't I tell you that uh that is not just going to be retreading the same old ground and it's actually interesting I think we're going to start with something that I see as most useful for everybody in the in the industry today and that is not specifically talking about what to do with AI but rather what it is doing with us uh because uh I've had a lot of these conversations you know I've been running an Nai company for almost five years now in in the games industry working with some of the best studios in the world some amazing AAA developers some amazing Indie developers and uh you can't help but feel like part of the reaction that a lot of people get to this technology

is this and believe it or not I I empathize I actually agree and it concerns me and this is exactly the same thing I've been talking about for 10 years and so the interesting thing here that I feel like ah we hopefully can acknowledge by the end of it is that the reason a lot of people have this feeling is because people like me who have been up on these stages you know for 10 years now talking about art and education have now done a good enough job have not actually gotten to the point of what art is what game development is why we do these things this is a 10 year old slide that was basically talking about how in our industry we conflate ourselves with technology we conflate ourselves with our tools we personally invest our identity not into the art but the craft and uh and it's very dangerous as it was dangerous for me as an artist you know starting whatever 15 years ago right where uh

I I did not realize where value in my work was and it's kind of scary if you think about it think you know 500 years ago you want to be an artist you're saying I'm going to go to school to study pencils people would think it's ridiculous today you're saying I'm going to be a 3D modeler I'm going to go to school to study unreal or Maya and it makes perfect sense but it doesn't the problem is that none of these things actually correlate with the with what you do and you're not your tools but also your tools are not your art and I feel like because of how complex the industry is because of you know how expedient is built to educational industry but also the software industry to make you think that their software is all there is to the profession it's very easy to get lost in that and think oh yeah this is well whatever the tools that I learn is what I am but that is not the case Just Like Music composing music has not a lot to do with actually playing an instrument sure you need an instrument to play music but you can learn to play an instrument perfectly and be an absolute dog shed composer and

what we need in this industry is people that are creative rather than recreative we don't just want you to recreate the same thing over and over we don't just want you to do things that somebody else has done we want you to compose and uh and I feel like this is not you know common knowledge we're still exist in the industry where a lot of Junior artists that that we hire you know just come in thinking that knowing which buttons to press and which software it makes them a professional makes them an artist or you know an engineer or whatever else uh and uh and that is scary uh but then the other part of it that is interesting to me and disheartening and you know that was another talk from 2018 uh where I was railing against that is actually how we also have gotten us so fragmented in how we work God and so micro specialized in our professions but that's how we've come to see ourselves within our companies within our organizations but also in our relationship to the final product in our relationship to our audience in our relationships two games to the the very

you know thing that we love and uh it's not uncommon you know to go to a studio and uh and basically see that there's people you know working on a portfolio piece more than on a game you know that's the conversation you would have where people are trying to make their part work better in spite of the overall product and I feel like that is also the result of us not acknowledging that if we want to have a creative inclusive thriving industry that everybody should be the owner of the overall experience we want everyone's goals to be aligned with the goals of the project we want every artist every engineer every animator to not just do the best job at their thing but the best job out their thing that makes the whole product better and uh and we're not doing that because once again the industry has gotten so big uh and uh it's hard to to maintain that but also it's hard to want to focus on it when you just have to get the product out the door and so uh yeah when I think about that I can't help but think of this guy but because I

imagine you know if you're like a drummer in the military in the 1900s and then all of a sudden someone was like you know what we kind of don't need drummers anymore you'd probably be not as upset because you don't want to die but to me that's like an interesting metaphor of like we understand in some industry Industries very implicitly that what we deliver to people is all that matters the result that we accomplish is all that matters yet somehow in the gaming industry we get caught up in I just like doing things that way so why can't I just keep doing that the answer is very simple it's because of your audience it's because of the people that you came there to serve in the first place because you're there to make a game rather than just make a piece of art and I feel like we all need a little bit more focus on that and so uh from having had countless conversations something that I find these days is that the fear of artificial intelligence is quite linearly correlated with experience and uh and I feel like that feeds back to some of the points that we just said and there's obviously a bit caveat about the ethical ethically sourced uh models

we're going to talk about that a little bit later uh but I feel like it's important uh and it doesn't just apply to AI honestly it applies to all the tools it applies to mocap it applies to but you name it we've been through this many cycles but I think the important thing that I keep repeating about tools and I'll you know die on that Hill is that tools don't raise the bar they raise the floor meaning you know when something like metahuman comes out how does that impact the people that are working their ass off to raise the bar and the question most of the time is usually it doesn't you know the people that are interested in advancing the state of the art are at any point in time very invested in uh you know the specificity the the very particular you know last two three four percent that just pushes them over the edge and kind of gives them a leg up over everybody and so whenever a new technology comes up it is not not disruptive to those people as disruptive to the people that could have not accomplished that without it and and now they have to transition and so uh yeah the interesting uh part

in all of that is just going back to that like hyper specificity right because if I can get that meta human that is 95 close to what my very specific design is it'll be great for an indie game if I cannot you know make something as specific myself it will not do for a triple a project where I have to be exactly you know the character has these traits these facial features cards where they need to be as well as special topology and things like that and so when we're talking about this technology and why senior uh developers are not as faced by it is because they realize that they live in that sort of hyper-specific last five percent of quality world where they have to spend most of their thinking and so this is just another callback to a presentation I would recommend from 2017 but basically most of the creative process boils down to how can I tell a unique story right if you take a long enough time frame where quality like for real quality is achievable for free to everybody which

is inevitable you know give it 10 years maybe even less what where does that leave us and that leaves us at a place where the only value that you have for your audience that is differentiated is some kind of unique purpose or meaning that you assign to that and so that would be you know the specific facial features of Joel but also literally anything else and so that is where most of the seniority kind of leads your brain over time is that the differentiated value I provide is not just in having something it's in having something very specific and uh and I talk about this a lot like people would ask me what is the difference between you know make being an artist and being an art director and uh and it's usually just that that bottom part right is most artists spend their time uh between the like zero to 75 space just trying to get it you know somewhere where it's gonna look good enough to ship in a pinch uh but the trick is is that the differentiation that happens there the thing that elevates your project you

know hopefully Head and Shoulders Above the Rest happens in these last few percent it's where you spend most of your mental capacities where you spend most of your time and to give you a better visualization this is kind of a rough graph for something that people refer to as the Pareto Optimum you can look it up really old thing from like 16th century but basically the idea is quite straightforward is that 20 of the work takes eighty percent of the time and vice versa so 80 of the work takes 20 of the time and what that means is uh it's very easy to get something to a baseline where everybody else can do it and we you know we see it in the industry all the time like say making levels of naughty dog right would be okay here's the Baseline we know that this looks okay and then how do you actually get it to a point where no one has ever seen before when it comes to Quality and and this is where we spend most of our time thinking about it right like the the Baseline the the block out is quick and then that creative part at the end is hard and that sort of creative and recreative window is basically defined by how many people can accomplish it right if if 80 of the population can do this thing then

inherently it has very little value if 80 of the game that has you know shift have this this level of quality then it has very little value to our audience because we're not differentiating and so the interesting thing is that over time senior artists lead artists uh you know uh they shift into that like how do we get that extra two percent you know we're always fascinated we're looking at characters and we all get you know super excited when we figured out that we can have you know knuckle blend maps that show like the bones showing through and when you Flex or something like look at the tiniest little thing that gives you that one more percent of realism and and that is the exciting place we want to be right when you got into this industry the goal is to Delight the goal is to entertain the goal was to raise the bar to go was the to hopefully make something exciting and so the question is if we're spending all of our time in that recreative space of just doing what everybody else is doing then why is that exciting to us why why are we attached to that part why are we so excited to just keep doing that instead of constantly asking ourselves how do we raise the bar and uh yeah and I feel

like that is that is the core question in art it's quite straightforward and this is once again just me bitching for 10 years in a row now but it takes in my experience an artist like four to five years into their career to start caring about lighting and cinematography and composition and this is literally the most useful skill you can have and I can't tell you how many uh an hours I spend educating people on these on these core fundamentals it's not about the modeling and texturing it's uh you know can you pick out a good color palette uh all of these things that are are the art you know that are not the tool that were that we have been missing for the longest time because we have been not applying the right focus and so to shift it to other professions as well because we don't want to just make it about art the the actually the long term of this is very simple we have creative problems we have ideas on how to solve them and then we have implementation time and so uh there's one thing that is absolutely certain and inevitable and that is that execution slash implementation time is

going to keep shrinking indefinitely until it's absolutely gone because that's what technology does and the hard part and the exciting part is coming up with the solution that is right for you that works for your project that works for your audience that adds unique value and that is what I always encourage you know my my teams to do whoever I work with it's it's not about the execution is can you actually come up with something that that elevates and so I guess the point that I'm trying to make here is that we are missing some very vital core comprehension a lot of the time that I find to these conversations about where the value is in our work about what is the difference between our tools and our craft and our art and at the same time what companies and people expect us to focus on as far as the value that we deliver to our audiences and uh yeah and so when people rightfully feel like you know their dreams are being stolen by AI I don't disagree I I can see that feeling and I

feel like that you know that conversation should occur with compassion precisely because of this because I don't think that anybody who feels uh betrayed in in this situation is at fault here you know they followed an industry they followed an educational system that conflated technology with with art that can uh you know that does not place the values in the right place a lot of the time and when you are invested into it when you're deep in debt because of it and then somebody comes up to you and says hey you know what this was not a value whatsoever you're going to be pissed too and so I feel like ah there's a few things we can do uh you know if if we are to progress forward uh with uh with some dignity here and uh and I feel like it's up to us as in everyone us you know the people that pontificate from stage and talk about education but at the same time people accompanies people in studios

people that you manage but also people that manage you and you know having having these conversations about what do our values mean for the for the people that we work with and how do they reflect on their sort of internal rewards functions their goals their fulfillment at work and I feel like these few very simple things should hopefully helped if we just start there and one is very simple we're all game developers first because if you're going into this industry thinking oh I'm just an engineer and I don't care what I do I just you know want to write full code more often than not you're going to run into a thing where your code is uh is going to be what's the word less than perfect because you have to ship a game and you know there's going to be Band-Aid is Band-Aids upon you know patches upon whatever else and this is going to be all duct tape together it's not going to be great because the code is not the product it's the the game that's the product it's what you ship that's the product same with your art same with your animations whatever it is that you're doing it's serving the game and uh and if you don't want to have a disappointing career I feel like uh it's

just a reasonable thing to uh to always keep in mind but I also feel like that applies to uh to companies as well meaning when we work with other people and we pigeonhole them into a micro sort of position where they're you know you're the weapons artist now or you're the guy that makes car wheels for a year now for the next GTA uh it's hard for them to feel like they're part of the process it's hard for them to feel like they're actually contributing to a game because of how we structure this you know we fragment and segment and we remove creative ownership and we put layers of hierarchy over people uh and then it's understandable when they feel like this little domain is all they have that their entire job is just like no I'm just literally the sculpting guy for rocks or something and then you take it away from you and all of a sudden I've been robbed and that doesn't have to be the case that doesn't have to be the case because fundamentally everybody should be thinking how did they make your game better how do they contribute to that and when they do then they'll just be cool what else can they sculpt or do I need to sculpt should I do something else you know it's uh when

everyone is invited to have that creative input you'll find that your product is better and at the same time people are more happy and fulfilled and so I feel like it's a it's a good value to stand up for the other thing that is kind of aligned with that is the idea that we don't all have to be in that you know bottom the recreation space of just kind of you know dredging away trying to do something that on balance many other people in the industry can do how do we all get to a place where we're doing something that is differentiated that is unique to us unique to our project unique to our team something that raises the bar something that makes our audience come to us why are we not spending 99 of our time on that rather than doing something you know that is kind of a generic part of the process and so yeah I feel like not enough people are invited to that thinking not invited to consider how are they raising the bar how we can do something that no one has ever done before and I feel like if people have their creative ownership their input is

welcome it's easier to feel like you're a game developer before before anything else and then the last thing that just helps is uh just this just general reminder and belief that tools don't matter I feel like if we just internalize that it's going to make it so much easier because it switches our language to the right things to the right values and I feel like our values fundamentally will Define how do we go through this transition how do we help other people go through this transition how do we make sure that even though it's not your fault you know even though you're running a company you want to make a good game uh you want to employ some Ai and you reveal it very right to do that uh how do you make sure that people want to work with you when you do that how do you make sure that people are excited to join on that Adventure rather than feel like you're antagonistic to them like you're saying oh we don't need anyone because you need people and games are better when they're made in a group with all of our creative ideas pulled together bouncing off each other with Creative Energy and excitement and uh yeah and all of it

just goes back to all of these core points and so uh yeah uh something else by the way that I think applies both ways right we talk about uh trying to invite more people sort of to this creative table uh because obviously the people that are just starting out are the ones that are most distressed about AI but at the same time I feel like it also applies the other way around whatever you think your job is whatever you think your job is going to be think bigger it is bigger than you think it is you can do more than you think you can and next year and the year after that you'll be able to do even more so if you think you came to this industry to do UV on wraps and model level of detail meshes no just just a hard note you deserve better and if you are not going to internalize that that the industry is going to do it for you and it might be much more of a root of Awakening that uh than you think it is so consider that think about how big you're thinking and then think a little bigger and believe that you can do more and be inspired and excited but

also feel entitled to because there's no other future for this industry where more and more things are becoming accessible where that is not the case and so yeah take some of that responsibility onto you and I promise it's going to pay off in in so many interesting ways uh okay that was that was the introduction um but at the same time as being an introduction it was also a conclusion so that was an interesting way for me to structure your talk but I figured that might be a fun way to do it uh but now we're actually going to switch into a more practical gear so what I wanted to do here is explain to you why or how we should approach these things but at the same time talk a little bit about you know how to be AI curious in a way that doesn't feel threatening and uh and there's some ways that are very pragmatic like you know this this section is not going to have big hand wavy oh this thing is going to come out in the future and whatever do all of these no this is going to be very like

you should go home and just be able to do it and and I would highly recommend um so and there's two things we're going to talk about you hear you know about generative AI everywhere uh I'm not sure if it's a term analytically I I like it because it actually defines the problem space that it solves uh quite well and it is the the thing that is quote unquote non-threatening that is easy to engage with which means that you are all sitting on way too much data and uh and there's uh you know technology that can help you process that they can help you gain Clarity and insight and a very simple thing that everybody can get behind is uh is just this like something that we use at work at all times right whenever you have in a meeting a call use AI all of the platforms have that to annotate and create just text outputs for an entire meeting and then uh on top of that just use another AI to generate uh sort of summaries of that and if you're working on a team if you're working on a small team production is a great thing to help automate with uh

with the AI because nobody wants to do it especially in a small team one on a bigger team you know there's so so much information floating around that you can't ever make sense of it you can't ever follow everything you can't ever be in every meeting and so if instead of you know sitting through a recording of a 15-minute video you can just feed the transcripts to another AI get 20 point bullet list read through it and be like okay do I need to dig in further and see what happened there or not gonna save you a ton of time so a very banal example of how you know your life can be made easier with these tools that uh that I highly recommend and uh the other thing by the way that's very valuable with this is that it makes it searchable over time so if you ever need to go back to something just type it into slack you'll find it and so once again very very trivial highly recommend the next one and I'm going to try and actually do it live and we'll see what happens uh is just trying to learn meaning a lot of the time what I do these days is if I'm just you know going through a chunk of code and I'm like I wonder

what this function does feed it to charge apt and then just tell me like what is it can you please explain to me how this works and uh and that is a great way to uh Empower yourself to do more things to learn more things because a lot of the time we like to assume that things are outside of our domain outside of our purview and we just shouldn't bother but the truth is that the information is out there you know you can teach yourself how to code if you use Google long enough but this is a lot faster right and and if you analyze any information that comes your way and just say can you break it down to me like a Vibe all of a sudden you're that much smarter for it and you know every time somebody at work uses a phrase you didn't understand you know every time there's like oh working on a script working on a tool working on this you're gonna gonna be that much better for it and you can follow up right so here's a breakdown of how it works and then I can say does this look like it might cause bugs

and uh there he goes and it's very exciting and for me right as someone who does a lot of engineering at the same time manages a team of Engineers as well as a team of artists like probability for bugs is a big concern and so just being able to say oh can you tell me how to make it more robust more more simple is valuable and you get you know the shared wisdom of the people on the internet just giving you advice live pretty much and it's not always going to be correct uh wait uh can you refactor the function take in this into account and uh yeah and so the idea is you can ask questions you can be inquisitive you can learn and then you can ask for advice and Corrections on any kind of text right because you can be writing a script for your game and you're just writing a story about a character and you can say oh well uh does it sound like this character is being persuasive or more you know uh I don't know self-sacrifice you know something like

you can ask questions I guess that is kind of the core point of this that I highly recommend uh and yeah and so that's another way to you know uh use analytical AI to help you in your day-to-day now the the other one uh is going to try another another live demo and this actually has to do less with text and more with assets let me actually switch to this other project and uh and just try to demonstrate but the idea is very straightforward we all have giant projects full of content right like I just downloaded a new asset pack it has a ton of stuff here I have no idea what it is I have no idea what's in here in fact let me run this real quick and so what you can use AI for that's very handy is to actually go through your content as well it doesn't have to build content for you but it can analyze your content and tell you exactly uh what's in there okay so the library the library is empty I'm just going to go to Unreal oh select all of

these assets and I'm just gonna ask uh our tools to add them to the library and so whatever it is 30 seconds we're going to process all of them and then all of your assets can be discoverable you know with just natural language after that which is a simple trivial thing right like I can't tell you how many conversations I've been a part of that were like almost a freaking knife fight at a game Studio like which folder did we put a thing into I'm sure you've been there and uh and the reality of it is is that we don't have to care it's 2023. sure naming conventions are only good if everybody memorizes them nobody ever does and so the very simple idea of just like well you know what we can just uh uh feed all of our content to Ai and uh and it just does it for us and so I believe this should be finished now and I can jump into uh the browser to demonstrate that for you and so that's it right these are all of your assets but you can at any point go in and say you know what show me food

right there's well there's food stuff show me foliage show me Shields show me and but by the way none of it has to be labeled right banners uh I don't know uh sharp things uh so any kind of language uh these are not too sharp or just like stuff that visually looks similar right so the idea that your content can easily be natural language recognizable is a solved problem right like it's it's something that is already here you can literally just prometheani.com download it for free we don't even charge for that and then you're just good to go and so the fact that you can use AI to save you a ton of time across potentially terabytes of assets that are just light laying around for you is a is a very simple uh is a very simple non-threatening solution and by the way it applies to learning as well so when we just did you know learning from uh text essentially and providing sort of uh providing uh I guess descriptions using Chad GPT

same deal we can drop in a whole bunch of visual content that is unstructured and this is just a bunch of last list screenshots and then uh the internet is not going to fail us let's go uh we'll be able to do the same thing ask questions but now visually about images about content and so once again 600 images in what 40 seconds and I can just go in and say show me cool examples of establishing shots right and then for everything I'm going to go in and say oh well these look like interest in establishing shots what about uh good golden hour lighting right it's like oh well that looks interesting okay and so you can have requests and then search and learn and then you can say okay what can we do with the uh close up shot of two women right and that is going to be this and so the idea that you can request things that you're curious about throughout all of your content whether it's all of your reference whether it's all of your 3D

assets 2D assets it's uh it's just there and so when we're talking about sort of non-threatening analytical AI they can just feed you information once again solve problem like there's it's not something that humans have to have to worry about uh at this point uh so highly recommend now we're going to talk about generative Ai and that is an interesting subject uh because I still feel like there's one more thing to be said about how we think about it uh and that is actually this and can anybody tell me what it is that we're looking at uh I'll give you a hint artists are terrified of it bow blank screen blank canvas exactly and uh it's true when I started doing watercolors what 20 25 years ago uh one of the first exercises that we did that was fantastic is you just splatter some watercolors on a page and then you just sit there and look at it until you see something you look for art within the splotches that you put on the page and it's uh and it's an exercise that we've

done in the games industry and concept art for for the longest time from photobashing to I don't know if anybody remembers the software called Alchemy from like 10 years ago it would just have random shapes that you could splatter on a page that then you look for designs in so you start with something you always jump start your imagination you're never at a blank slate and it's not a new thing we've been doing this forever and uh and I feel like that is the right way to look at generative AI is that we don't have to start everything from scratch at all times and think about movies right every time they make a new movie they don't just go well let's make all the furniture from scratch you know you walk into coffee shop say cool change the drapes you know put the camera here lights there we're good to go and that's how it should be whatever you're doing if you have somewhere to start you're going to save a ton of time and so uh that is the Hope with all of that is how do we give you a jump in a point and I feel like the important thing to note here and uh oh well actually yeah we'll get to that in a sec uh and so talking about jumping off points right uh just

just brainstorm right whatever you're working on is like oh Chad GPT can you give me 10 ideas for a cool environment portfolio piece to get me a job with blizzard and then just just Grill it like oh what what what is good about this idea like actually interview it like it's your assistant and just come up with interesting ways to uh interesting projects to work on and you can do this for your game too give me 10 cool feature ideas for an indie game you know and then just interrogate it like oh maybe it's whatever jumping mechanic how can we put a Twist on it give your brain the same you know watercolor splotches to start with instead of just sitting there in a vacuum because that's what you would usually do right you would go look for inspiration you'd play through 25 other games you go watch a movie this is essentially that but you're just also getting the collective wisdom of the internet for reflecting back your questions at you which is of a lot of value and so yeah whatever idea you have run it run it by uh this is just for me personally there's things that I just don't like doing like uh social media and so these days I'm just like oh we need to write a tweet it's like I'm not

writing a tweet I'm gonna get a chat GPT to write a tweet and uh it helps and so that's the other part something that I'm noticing about technology that is actually being quite empowering is that it doesn't just have to focus on your core expertise if anything it's fantastic at the things that you don't want to do that are not your core expertise because those are the ones usually procrastinate most on and you're just in there like oh I need to get to this but you know maybe tomorrow next week because you don't like it well you know don't wait to delegate delegate we should all delegate a little more and so uh yeah I feel like having that train of thought for generative things that it doesn't just have to be the things that you like doing you can actually delegate the things you don't like doing is very valuable think about that more what is actually making you happy whether you're excited uh to do and then same with writing code right like a Promethean we have Integrations into every 3D editor and some of them's like not worked much with so every time we need a blender script it's like yeah no Chad GPT go and it's the same thing for

engineering right it's that the watercolor splotches where if you show me a chunk of code that works somewhere in a particular software I'll figure out how to change it and adapt it and make it my own it's it's starting from the blank canvas that is killing me because I have to go around and you know search on the internet for 20 different examples and find which one works and all of that stuff and so uh yeah and the cool thing by the way that I highly recommend for all the game studios right the big problem is we have a lot of Engineers and then nobody's helping artists because that's kind of very specific you need technical artists just this would like how to write scripts in Maya plus any one of your engineers in your studio and they're a technical artist that's it like that is The Interpreter and so there's a lot more opportunity to help people in sort of these neighboring uh kind of uh positions or Specialties just through using AI As an interpreter mechanism and I highly suggest you do it and it goes other ways you know it's like Engineers can do more code but at the same time I see artists making their own scripts now with this because why the hell not you want to save time making art so just just delegate uh and then the domains you are familiar

with which is interesting so uh we at Promethean now try to do at least 10 of our code with AI whenever we can and uh and it works the interesting thing is it's just faster and so the interesting thing for us right is asking the right questions getting the right context building the right features because you know you don't want to be stuck at that level of I'm just thinking about you know the the letters and the and the function names how about we try a feature and then we throw it out if it doesn't work like that's an amazing flexibility to have every product is defined fundamentally in its quality by the ability to iterate and if you can iterate faster that's how you get better products and so all of our technology has been kind of working early there and eventually yeah we're just going to iterate whole features we just request a feature which is uh which is very exciting we're going to talk about 2D a little bit but I think this is mostly going to be some useful caveats meaning you know we can just go online run any of these and uh and get some

results uh but the one thing that we want to start with right is just the legal liabilities of using that today and uh I'm sure as many of you have heard there's plenty of lawsuits going around uh class action lawsuit from a whole bunch of artists there's a Shutterstock lawsuit I believe against table diffusion as well and so there's a plenty uh plenty still to be decided I've just full disclosure uh I'm an expert witness for the artists in that case and so it's very interesting being a generative AI CEO trying to help artists Sue generative AI but I do believe that we need a very ethical industry that is concerned with the artist's well-being that empowers them the right way that doesn't abuse their data or their intellectual property or even more so it feels entitled to it and there's plenty of ways to do it right and what we need to do is set the right precedence set the right examples set the right regulation and then make sure that everyone who is assisted by these

workflows is also benefiting from them rather than uh rather than getting fully supplanted and so that's kind of the point that I was trying to make before uh you know a generative AI in general is uh if the output of this is 100 of what you need meaning you think your job is done just because you typed in a string of text that means you don't have deep enough domain expertise about what you're doing it means about it means that that specificity that we talked about at the end that curve and the last four percent or five percent it means that you've not gone there mentally because the quality of your product is defined by how well you apply it to your story to your specific intent design idea it's all about just making it very specific and I'll give you an example right like Uncharted 4 we're making the level where Drake is like washed up on an island after his boat is crashed nothing is accidental the choice of colors the choice of shapes you know we have foliage across the game that is very varied that one section where it's the characters at

their lowest every plant is as pointy as it can be every rock is as jaggedy as it can be every color is as cold as it can be every surface is as wet and miserable and muddy as it can be and so the idea is that we assign meaning that is the value of what we do we take the creative intent and then we express it through every single little detail in uh in in our work and so when you're generating this and if you just put in a prompt and you got a result that is immediately satisfactory then I I probably can vouch that you've not thought deep enough about what is the meaning that you're trying to convey and and how are you doing that and so this is great for ideation this is great for references and concept art and you'll find that it has a lot of gaps uh is as to just how it understands things and it knows some things better than others but two things to note here one is that there's a non-zero chance that any one of these uh diffusion models you know traced on that train or

that lion data set can output a piece of art that would be considered plagiarized if a human put it in a video game meaning uh it's not guaranteed that it will do it every time but there is a non-zero chance and what that means is technically if you want to get rich quick by the way you've heard it here first find some game that uses user generated content using that stuff generate a thousand little images as examples and then use reverse image search in the data set to find if it actually looks like something that someone has already made reach out to that artist get him to sue the company that made the game make a lot of money and so uh but at some point it's inevitable and that is the tricky thing we're trying to make tools for commercial projects if there's legal liability right like a Sony for example would triple check every piece of media that goes into your game because if it's someone else's intellectual property that means that you know the 10 million Blu-rays that you've printed now have to go off the shelf because you can't ship up people too they put an injunction on your content and so it's not a safe thing to do and until that is fully resolved

uh most big companies have that kind of regulation in place right now and I would also recommend you as well to not use this as final content until you can hear and know about how this problem is resolved but at the same time the other part is just ethical and it uh kind of honestly inspires me and warms my heart sometimes I talk to Studios and they have the legal go ahead to use some of these tools but they're just saying you know what we're gonna wait we're gonna wait and see how the lawsuits pan out because we want to make sure that we're inclusive and kind to the creative community and if we can't afford it we're gonna do it and so if you can't afford it that's great if you can't afford it I can't say I would judge you uh but yeah that is kind of where where that is today uh the interesting thing is obviously everyone's talking about 3D so I figured I'd give you a quick overview because you know we run Tesla permit then all the time the moment Point e came out we ran it through like 3000 different object permutations and the interesting thing with 3D is that you know it's still a little bit far off can can anybody guess what this is

yay it's a flower bouquet um but you know the other thing that is interesting too uh to kind of visualize here is just the depth of the latent space meaning I tried given the same prompt to 3D generation and uh 2D generation just to see the difference and you can uh you can kind of tell uh but the other thing that I don't feel like we consider enough right is that going from 2D to 3D is not just adding an x-ray Dimension I mean that's a step but fundamentally there's so much more resolution that is required and then there's layers of material information that is required on top of it for every you know 3D Pixel you need to know your roughness you know specularity normal information Albedo information subsurface potentially also transparency if you have that and so the actual amount of Dimensions is not going from two to three is going from two to ten and so uh there's plenty of time left until you know this gets properly solved so there's uh I'd say two to four years

there but the interesting thing to me and that is this is more coming from a lot of like Tech conversations that that I've been having with people and even on the investment side right so there's money going into investing into solving this uh it's interesting because especially in the AAA space I would say for the last decade very few companies have made assets in-house and uh and it uh checks out in a sense that if you're trying to make a game if you're old game makers there and you're trying to focus on what you know uniquely provides value what you do is you try to define a vision you try to define a style and then these assets can be produced much quicker and cheaper somewhere else and so just the same way you know we had this conversation that we're having about AI you know 10 15 years ago when we were talking about Outsourcing it's it's very much a thing right now and so the other interesting part when there's conversations about disrupting sort of 3D asset production I'm not sure at least in the AAA space where is that disruption to to be made because most of them have not been produced in-house for the longest time

and so that's the other thing actually when it comes to education and I've been banging that drum for a good eight years now there's no such thing as prop artists in you know inside of game studios the value that we that we create is we assemble worlds we tell stories we drive the player through the experience and then the props are just made elsewhere but the other thing that is interesting is that on a long enough time frame just like when we're making a movie right the whole world is going to be scanned in sound capacity and so just like you know Ridley Scott does not go okay we need to make all of these chairs by hand right now before we can shoot this movie you're also not going to be doing that uh you're going to be assembling you're going to be doing cinematography you're going to be setting up lights uh you're going to be yeah just doing all of this cool stuff and so that's 3D and then yeah for the uh for the final part of it I actually just wanted to give you a little demo of of generative AI in a way where it can uh be inclusive and hopefully inspiring uh and uh it's actually going to start with this and so

just to give you an example because obviously we've been thinking about that problem for what five years now ah and uh the way we try to achieve that is by making sure that the people that are getting the benefits of improved performance are the ones that are the authors of their own Automation and that's kind of the core thesis there is that we've been working for years with this concept that right there's people with answers meaning engineers and there's people with problems everybody else for whom they're making tools and uh but there's this giant Gap in the Middle where we're just you know literally throwing pits throwing like trucks of money into that pit every year because communication is hard and so wouldn't it be great if for example an artist could make their own tools that automate their art and maybe a designer could make their own tools that automate their design without having to know how to code and the exciting thing about AI conceptually is that it's meant to learn from examples that is the whole point and so for example when we train World building with Promethean right that is what it looks like this is a scene that programs it to build forests and then

you have sort of an example layout and then you have some some little individual trees with compositions and then all you have to do is just arrange it and then literally say memorize scene and then it'll just go to work and memorize this and the next time we want to build the forest and we'll build it and so that is uh what we're going to do right now and so what I'm gonna try to do live is just illustrate a little bit of that workflow okay we're not going to do that we have a scene here that doesn't have any lighting so let's say load uh dusk lighting real quick so that we can see stuff nice uh okay maybe we go unlit a little bit so we're trying to do just let's make a little vignette and uh do this high level creative problem solving so that's kind of how we refer to it internally right is uh I want high level creative intent turned into uh you know final 3D content pretty much and so

the idea here is I'm just doing a little outline with the perimeter tools and there's going to be enough for me to say you know what that's where I want a forest to be right so we just saw uh what a forest would look like and by the way going full screen is kind of our thing where like the whole point of doing art is that you shouldn't have to look at 28 sliders so now we can just say build a Northern Forest inside of this and uh hit enter and it should go to work uh oh that's right way I need to switch projects never mind uh let me do that real quick while it's well it's grunting here uh oh no we're all project live demos don't you just love them and uh okay okay still building uh

switch project to Unreal five let's go all right I think we're getting better now yep finished generating all of this stuff as you can tell without the assets because it's not connected to the assets and I'm going to go back and do it again if it's going to look good uh okay go go so now we have our object so now we're gonna uh build a force here and uh yeah we have our star to Forest and then we're going to do a little thinking and then uh more smaller assets are going to come in so you can tell there's a bit of a progress bar but I reckon you then in under six minutes we'll probably have something that you could ship in a game uh as a level so bear with me uh okay so uh I think we're gonna do right

is we have this little patch of land we have a little Forest back here we can set up like a little Farmhouse and we're gonna do sort of a stranger things inspired like kids get lost in the woods uh type deal which I think is going to be fun okay 62 for the forest almost there and uh what you're gonna see right now is just all of these objects that you saw trained in that training level like all the small little detail yep and be dropped into the level and so now if you look around it's just fully populated with little tree trunks and rocks and whatever else you have and so uh yeah you can ship it in a pinch and then from here we can just start picking out compositions so maybe we want something like this uh let's remove this guy to give us a little room uh I'm gonna bookmark this camera and then uh let's build more right so we said we want to farm uh we can just ask for me again to give us Farm plot objects right like what goes into a farm

and uh it will just give you a list of like oh you're building a farm try these right and it's like here's a farmhouse here's a gazebo Greenhouse storage whatever it is and uh by the way if you unclick it this is to the idea of like analytics here's everything you can put in a farm from your library just pick something if you like it and then if you want kind of variations of this uh let's say you're gonna go up here maybe I want a different Farmhouse I could right click here and you know see all the other farm houses that fit but no drag and drop it into the world and just put it to work and so I'm gonna wait for unreal to add some stuff Oop There we go and uh yeah we have not just a farmhouse but a farmhouse that's like dressed up with a whole bunch of stuff uh oops I'm gonna have all of the objects here some objects here and so not have to spend time doing that at all looks nice okay okay let's get in there see um it's like a little composition is

coming up maybe do this and then let's just add a little more grass here and then we have our focal point and we're pretty much done with this uh so let me just switch to this uh up I'm gonna add some grass here so it looks nice and detailed up and uh to build a meadow inside now you do have to know how to type oh exactly the irony okay it's working let's do this thing come on buddy oh all right okay so this will be a little patch of grass

nice and so this is our view getting there uh it's gonna be doing a little thinking right now but you know what we can search for a car to do like a little abandoned thing uh well it's still doing some thinking let's drop it into the world here's our car and uh by the way if we're very lazy unreal is grunting let's get in there if we're very lazy we can do things like face the house and then Promethean should just rotate the car to face the house uh oh wait I couldn't find the house wait let me try that again uh oh well uh I guess uh too many Optics on screen sometimes that happens never mind bear with me uh okay so we did we did this I feel like we're quite close uh you know what let's grab actually this tree trunk here just move it up here blocking the road uh and then maybe add a couple of characters oh you know what we can add we have like scary birds

uh I'm just gonna search for Birds uh it's like a particle system place in here and then just search for a few characters and uh search for ones with a backpack right and I'm gonna go in grab this guy and maybe grab this guy and then they're maybe waiting on this guy uh cool so now I'm just gonna turn them around uh and uh yeah I feel like we're pretty close and so you can tell high-level creative intent to pretty much a full environment and actually you know what let's try a different lighting and see what happens slow Dawn lighting uh

and uh yeah you know a little Children of the Corn Vibe but uh I'll take it and uh yeah I think we can stop here I'm not going to detail it up much more but I guess my my point of it was that if we do AI write right if we build technology work I'm an artist I can just give it a few examples and say make more stuff like this and then we have a technology that is caring about creativity it's caring about including your voice into it in five minutes you can build things that would take you a week to do before and what that means is you can tell more stories you can tell more interesting stories something that we've done at Promethean a couple of years ago we made a virtual reality experience about and this was actually my GDC talk last year going back full circle for referencing all of the talks uh a story a virtual reality story about a five-year-old girl that got hurt in Belarus when police was indiscriminately attacking people and they smashed into her car with her

parents they just crashed into her car and uh yeah and the idea that we can tell that story now because we have technology that can allow us to build Virtual Worlds you know literally in hours that are shippable 3D products is fascinating to me and the thing that I would like to see is more people feeling like not just included but entitled to telling these stories to saying hey you know what we can do more with this uh we don't just have to be worried about the recreation aspect we can be worried about creativity we can be worried about making experiences that were not possible without this and uh and yeah I guess fundamentally going back to just uh the that slide back right when we talk about a lot of people feeling like their dreams being stolen wholeheartedly understand that and I feel like all these values that we talk about to include about including people are Paramount in helping them not feel that way but then after that is the next step which is that we actually get to think back about those dreams what were

your dreams you know if uh if they're being stolen and uh if your dreams have anything to do with making amazing video games then you'll find that you actually get a lot more back but you can do a lot more things that you couldn't do before you can build big amazing worlds and uh and tell stories that have never been told before and I hope that that's fundamentally what we're going to do we're going to take care of each other we're going to bring up the right values and set the right examples and make sure that everyone who has unfortunately been bamboozled into a situation where right now they feel you know uh robbed of their dreams that we can take care of those people we can remind them there's a place for them we can remind them that together we can build amazing things and that they are destined for amazing things as well telling these big stories and uh yeah and hopefully we're all better off for it this is my talk thank you so much everyone if you have any questions feel free to

line up hello hey there I've got a question for you um learning how to solve big problems in my opinion requires solving smaller problems and this is one of the reasons like in math and many many problem domains so as a more senior person like I've already acquired that knowledge and I'm a fully comfortable solving probably big problems across many domains but I'm concerned that if like an AI system is solving all the small problems it kind of like pulls the ladder up behind me denying less less Senior People the opportunity to solve those smaller problems and so I'm curious if you see a solution to that that is a great question and my solution to that would probably be wondering if if that is a problem or if we all just go up a step meaning what you take as bigger problem now hopefully are going to be small problems for the people that come after us and I feel like that is a good way to do it in the sense that we're not you know pulling up the letter behind us but

we're elevating the whole thing with us and if everyone should feel uncomfortable there it's not the junior artist but I feel like the senior artists asking ourselves what can we be doing more that we are not doing are we getting complacent with our skill set right now or should I just not feel feel like I'm good at making a level can I be a good art director why am I not a great art director and then why am I not a great game director why I'm not a great Storyteller all of these things just compound but I feel like it's a good way to utilize this technology and this approach right and just asking ourselves how can we be better ourselves rather than uh yeah rather than just for other people so hopefully that answers your question thank you hey really cool Tech I saw you're showing casing unreal is there any Unity integration uh there is it's not public but hit us up we can send you a hookup perfect awesome thank you I'm serious can this be integrated into in-house yes 100 full in-house Integrations plenty AAA Studios do that already

uh hey I'm a game game designer um like speed's really important to me so I might use some like AI stuff to like do documentation things but I still have to like bump into specialists in the kitchen any tips for like sort of explaining to them that I might use this and I still value them and all that sort of stuff oh that's a great question I feel like the trick is for them to have the same buy-in for like what they can use it for and uh it was actually interesting so we were on a panel yesterday with Kelly who's all the way back there and he talks a lot about uh uh making the like the tutorial videos uh and uh and it's kind of one of those things we're like oh if you're a specialist one thing you definitely don't like is explaining your thing to other people all the time so if they can make you know summaries uh and and use AI for that for example that's like a simple way to say oh hey I use the thing to know more about your thing but fundamentally right it's just the buying for everybody like these things are so fundamental uh but the other interesting thing actually that I find fascinating is learning more about other people's

field of expertise through that and then asking them what would they find you know extremely important about their space and what do they think is sort of inalienable like one of the talks from 2016 or 2017 had the slide that the whole cool thing about AI is that it forces us to have a conversation about what art really is what design really is what animation really is because the thing that we refuse to give up is going to be the thing that defines it in so many ways and so having these conversations with people about what would you refuse to give up I feel is very interesting because it shows you where they see the value and what they do great thank you that's helpful yeah hi amazing talk I mean it resonates like with all with so many things in me and uh I I had adjusted one thought bump up like when you said that uh senior artists should be uh wondering like when the new generation grows up they will have all these tools and uh

what I think is that everyone should use them to um increase their productivity but the junior ones are actually ones who are not like endangered as they will be the first to adopt it because they're like the the youngest one so so no worries there I would say that's a fair point I think it's a double a sword in a sense that some people see it as solving their job entirely meaning like oh I was just going to paint concept art and now I mean Journey does and they're like oh then you're missing the point of concept art because it's actually quite specific to the problem that you're solving and people pay uh not for pretty pictures but people pay for solutions to problems when they do concept art so I agree with you wholeheartedly that they might actually adopt it but yeah from my personal experience just talking to people there's definitely a lot less trepidation with senior people somehow like the more senior people get the more they're like no it's great bring it on you know and then but sometimes you also get very callous you know veteran are directors that are like this is inevitable deal with it and I'm also kind of cautioning against that in the

sense that it may or may not be inevitable it doesn't mean that we need to lose compassion and understanding for the people that ended up in the predicament that makes them feel like you know it derailed their entire life and their plans for a career and so we can be sensitive to that and help bring people back in and welcome them back into into this industry thank you okay hi uh from my understanding it seems that a lot of the jobs that people get to kind of like break into and get a foothold in the industry seems to kind of live in that recreative space which seems to be the most impacted by AI as someone graduating in two months who's trying to who is focused on environment and prop art I'm curious if you have any uh recommendations for someone trying to kind of get into the industry in this strange transitionary time I do actually and I feel like they're mostly the same which is uh as much as you can try to not think of yourself as a junior person trying to

break in if you can look at a project and say how can I bring value to that that is the only thing that matters and that value I find that so much of the time working with less experienced artists Engineers whoever it is their main metric for quality or sort of contribution is what the boss says right it's like it's trying to please the person that's supposed to take the responsibility for the quality of your work but you don't have to do that in fact we would much rather hire people that can just look at our game and then make something that kind of looks like it like that's that's all it takes and then just just figure it out like can you be thinking about the quality of the product at all times so that I know that we are aligned instead of you just thinking of how to please me or follow to a t whatever I say and I feel like that alone is so rare and so Priceless in in hiring people because fundamentally that's all you want you don't care uh for the project you want autonomy and then some basic degree of Competency where I can train you up but so then talking about like how do you break in look at the game that you love and then just move reimagine a chunk of it just be like hey if you guys were to make a DLC here's what it looks like

here's an idea you know and uh yeah and use everything at your disposal and just be like hey I can I know what good art is because fundamentally by the way when you're doing a portfolio right you're evaluated on your worst piece at all times because we're not looking at canvas guy models like does he know what pretty is does he know what is shippable and what isn't because if you know that you'll learn everything else if you don't then it means that it's just going to be a long cycle of back and forth of like oh uh is this good we're like no do this is this good no do this so whatever you do just demonstrate that you understand what good means for that project and you'll be you'll do great I don't think you're going to be handicapped in any way but these AI tools thank you I'm just wondering and you you said we should emphasize on the creative part of of us but the thing is we want to create something that would be appreciated and the thing is AI actually I mean by collecting Big Data they actually knows those players probably even better than

us so how how can we like trying to that's actually a great question there's multiple answers to that but there's one big thing and that is that artificial intelligence Technologies by definition generalized meaning you can take you know seven billion people all of their data and then you ask it one question it's going to give you one answer this does not happen in real life so all of the things that by definition are made by these tools are generic and then it's your job to figure out how do I make it unique how do I add a spin to it and that is kind of what you know the what I was talking about with the movie industry right where you might shoot a movie in a coffee shop like a thousand other movies have done before how do you make it your own can I change the drapes can I change the lighting can I change the color you look for things to change that are unique to your story to your narrative to your experience and I feel like that is a value but then there's this other very interesting thing that's been bugging me to no end and that is that you know when we make a triple A game it's like 100 to 200 million dollars to produce and then probably 200 million dollars on top of that to Market which means that we can only afford to tell stories that appeal to everybody

and their dog and uh and that means that we're locking ourselves out from actually much more Niche stories when you're saying how do I uh you know make things that people like you don't have to make things that everybody likes and with this technology it makes it a lot cheaper to do things you can make something that a smaller group of people will absolutely love and that is very exciting because that is where you take that generalized idea you use it as a sort of a backdrop and then you kind of you know make it your own make it almost niche in a sense but if it's you know if it cost you a million dollars to produce a niche Triple A quality game you can make a great living because there's going to be plenty of people that will you know buy it and so I feel like it was kind of two two ways of looking at it greater advice next thank you I think we're all done thank you so much folks appreciate you coming back [Music] thank you