Roblox CEO: We Want AI To Generate Full Games — With David Baszucki
51:47 Share

Roblox CEO Dave Bazzucchi joins us to talk about creating games with generative AI, safety on the platform, and whether you should give your kids money to play with on Roblox. That's coming up right after this. Want better ad performance? Rated number one in targeting on G2, StackAdapt provides access to premium audiences and top-performing ad placements. Visit go.stackadapt.com backslash LinkedIn and launch winning campaigns today.

I'm Tomer Cohen, LinkedIn's Chief Product Officer. If you're just as curious as I am about the way things are built, then tune in to my podcast, Building One. I speak with some of the best product builders out there. I've always been inspired by frustration. It came back to my own personal pinpoint. So we had to go out to farmers and convince them. Following that curiosity is a superpower. You have to be obsessed with the human condition. Listen to Building One on Apple Podcasts or wherever you get your podcasts.

Welcome to Big Technology Podcast, a show for cool-headed, nuanced conversation of the tech world and beyond. We're joined today by Roblox CEO, Dave Buzuki. He is the CEO of a very large app with 85 million daily active users across 190 countries. They're not all, uh,

They're not all children. 60% are over 13 years old. So we'll talk about the appeal of the platform, the safety measures the company is putting into place to keep kids safe, and

whether or not it's a smart idea to give kids money when they're on. Because I know my friends, they're kids, they're asking for money for Roblox, and I want to speak with Dave about whether they should give the kids that money. So mostly excited to speak about the AI stuff. We're going to go into depth about creating games with AI in the first half, and it's my pleasure to welcome Dave to the show. Dave, great to see you. Welcome to the show. Thank you so much. I'm really excited to be here, and these are fun topics. So thanks for having me.

Definitely. I am a user of Roblox. I can't say I'm an active user, but I do know that when I'm around my younger cousins or my friend's kids, I fire the app up and they show me how to play the games. And it's really, it's usually a pretty good time. Well, hey, welcome fellow Roblox user. I've been a Roblox user for 18 years going on 20. Okay, great. And so the question is,

What are games going to look like in the future? Because to this point, you have a platform where people can build games and then you could go out and play the games yourself. And a lot of that is pretty painstaking, taking certain features and moving them in certain directions to enable certain actions. And you could spend a lot of time building games. Building games takes a lot of time.

Then this thing called generative AI happens, and all of a sudden, the idea to build becomes different, where maybe some people say you could just prompt a game and it will show up. Now, I don't think Roblox is there yet, but I do know that it's on the roadmap. So talk a little bit about how you envision generative AI coming into your creation process and where things stand today.

you know it's pretty interesting if we go back 20 or 30 years and so many of the things um maybe we would imagine 20 or 30 or 40 years ago and we'd read them in sci-fi books some of them have started to come true the you know the dick tracy watch has started to come true and what we would think of as the web has started to come true

And when we started the company 20 years ago, we had the notion of this 3D digital platform in the cloud. And if we gave people this 3D platform, we would see interesting stuff. To zero in on your question, I think the fun answer is we don't even know. And I feel that the reason we don't even know is there's a nuanced angle to AI that

which accelerates all of us as humans, whether we're an artist or a 3D designer or just someone with an idea. We accelerate that. It helps us be a better creator, arguably just like Photoshop helped people who used to paint with oil be better creators. So we're going to see a lot of human accelerator. But there's a second nuance, which is

What is a game? And as we start to have, in addition to AI creation within the games themselves and within experiences, the ability for all of us to access AI at any point in any time, it creates almost a new opportunity where we don't know what a game is. And where I'm going with this is in addition to AI creation,

Pretty soon, we're going to have 3D generation and text generation available within any Roblox game. And so what that means is, for example, a fashion experience like Dressed to Impress, where we typically put things on and shop and put an outfit together, will be someday complemented with you or I dressing

describing what kind of outfit we would like to create. And so it's interesting to imagine that AI may accelerate creation, not just for game creators, but for every player of games. - Right, so if you want to customize your experience, maybe you can prompt your way into a character that looks exactly like you want it to look like. - And people who watch Project Runway and say, "I have those ideas in my head,"

I just don't know how to sew them. I don't know how to cut the fabric. When they start describing those ideas with text or voice prompt,

they may start to see those ideas come to fruition. That's just a really small vertical slice of why we think there's an emergent new world of what is a game as AI starts to power some of these experiences. How far is Roblox from a place where you could say, you know, imagine a world where, you know, the characters are dragons and I'm trying to do this and these are the objectives and it should be elegant. But, you

but easy for users to play, and then it just creates the game in and of itself. We're getting closer at remarkably high speed. And one of the accelerants of this-- and this is true for any AI system--

is the more data that's available for training in a legal copyright to s compliant way, the more quickly we can build up AI power and Roblox is unique in that we have an enormous amount of 3D data, 3D objects, 3D scenes. There's hundreds and hundreds of medieval castles. There's thousands of cars, there's thousands of dragons and all of this.

We're using that data to create not an LLM, but a 3D foundational model within our AI team. And we've said publicly, we're going to release the first version of it this quarter. And we're going to release it in several forms. We're going to release it

as a open source 3D foundational model for others to use. We're going to release it in Roblox Studio, so you will be able to say, make me a good dragon. But we're also going to release it for creators to use within their games. So we will be testing the very first version of you and I in a Roblox experience today.

saying you make your dragon and i make my lion and we we start to see them emerge obviously it's version one um but this is going to happen this quarter what about like building like i guess these are elements of a game um do you envision actually being able to prompt and play a game

I think this is really, um, insightful because elements ultimately add up to the full experience and, and the full experience of a game, especially on Roblox is, is really quite rich. We have 3d objects. Um,

Unbeknownst to a lot of people, the 3D objects on Roblox actually tend to be physically pretty realistic. We run physics simulation. Cars have wheels. If wheels fall off cars, the cars skid out. We have 3D terrain. We have a lot of code embedded in all of those objects.

Roblox game creators embed code at the world level and the object level. And you're exactly right. The North Star is not just object creation, but full-on experience creation to where one could imagine someone drawing a few sketches of characters, describing a few fun type of gameplay, and literally generating their own Roblox experience from that.

Can I tell you a crazy thing that I did? I was, I mean, crazy. Who knows? You be the judge. I was on Claude earlier today. I guess I was thinking about this interview and I said, can you create me like a game where a journalist hunts for scoops and they have to ask the characters for a scoop three times. And then at the third time they will give you a scoop and you get a point for each scoop.

And it actually builds the game, it codes up the game. And it starts out as like, it's just like a text game. So like you click on the character, and then you, you know, ask three times, and then eventually, like you might get the, you might get the experience built out. Then I prompted it again. And I said, actually, I want this to be a

a town square that you walk around where characters walk around. So it creates it and it has like depersonalized emojis. Like the lawyer is like a law icon. I said, no, I actually want these to look like people. And then it goes ahead and recodes it. And now then it works. Literally took me 10 minutes. I coded it up and even gave it a name. It was called Scoop Hunter, which I didn't even prompt it with.

Next thing I know, I'm going out and playing it. Now, I know that's much more basic than a Roblox game, but it does make you see the path to be able to get there. So I'm curious what you think about that and how far away you think we might be from being able to do that within Roblox. I don't think we're too far away. And I think what you're doing, it's fun to imagine the prompts you gave Cloud

to trying those same prompts in a fully immersive 3d environment, ultimately over time. And I feel where you're going with that is exactly right. I like the notion of your game scoop hunter. I like the idea of the, um,

the full physical fabric of that system is a high resolution 3D environment, a 3D environment that can run on any device, a 3D environment that can be multiplayer around the world. And that fabric then can support that query. We're going to need some NPCs powered by AI. We might need different personalities

that think about their scoops. We might need a great town square that you would stylize more as you discussed it. And that'll all come to life. So I think we're not that far away from that. And so not that next year?

I won't say when, but I will say we're going to ship in Q1 both text generation as well as 3D object generation in Roblox experiences. And we would philosophically agree with you that full experience generation is the highest way to think about that. A full experience that's immersive is 3D objects, it's prompts, it's code, it's pulling all of that together.

I think where this ultimately leads to, as computers get faster and have more and more performance, is almost where you're going with that is happening in real time as you're walking around a 3D environment. And the ultimate, when we think about how you and I might dream, when we're in a dream environment,

We're literally generating a 3D environment in real time that's changing very quickly as we move around that dream. And so in a way, what you're describing about if it's done in real time and you're walking around Scoop Hunter and you say, whoa, morph into this. I want sci-fi Scoop Hunter with the Supreme Council of 10,000 scoopers and it morphs into that. You can imagine all of that happening in real time.

So you're saying you want the players also to be able to prompt as they're there to change the game experience. I think the more we think of a game experience as available to the player, not just the traditional game creator, the more robust the technology is. And we, you know, at Roblox, Roblox Studio is a very powerful tool for

All of the facilities in Roblox Studio, we want available to the game creator, AI creation, interactive modeling, 3D things. So there's really kind of a blurring between the studio tool and the creator. And what you're talking about could be done by the user in real time. So just to confirm, this is your North Star. You're interested in...

Well, let's just say goal. This is your goal. You're interested in creating an experience where I could prompt an experience out of thin air like I did with Scoop Hunter. And then as I'm playing a game or I'm in an experience, I as the user can prompt my way through it to make it sort of shaped the way that I want to experience the experience. Yeah, and I would say it's bigger than a goal. I think there's a class of technology problems that don't involve...

things like faster than light travel or hacking human biology that we can imagine with more compute and better algorithms are naturally going to emerge. This is one of those that I believe is just naturally going to emerge. It gives us the opportunity not just to set that as an interesting goal for those who want that type of environment,

but also something that will naturally emerge. I'm not sure everyone 24/7 will be running around Roblox dynamically creating environments all the time, because many of us will be hanging out with our friends, going to a concert, many of us will like to go to the creations of other people, many of us will still like to play hide-and-go-seek, but

For those who want to use that type of experience or creation, I do think it'll become available. Yeah, because that is the thing about this, right? When you play a game, the thing about it that makes it so enjoyable is the taste and the talent of the game creator. That's correct. We like to play Assassin's Creed in this house, and part of the joy of Assassin's Creed is just the brilliance of...

the Ubisoft creators who have gone out and built that game. You experience it in the quests. You experience it everywhere. Now, when it comes to prompting a game, it's almost like you could have a quote-unquote normie try to prompt a game, but I'm not as bullish that they'll be able to do what the greats can do and replicate what makes playing a game so magical. It reminds me a little of...

When I was younger, I got a Mac Plus and the Mac just came out and the Mac had WYSIWYG windows and fonts. And all of a sudden everyone thought they could do text layout and typography and

and we saw all these documents with 39 different fonts that look like crap and i think there's a little bit of an analogy to that where font craftsmanship and layout used to be done with movable type it moved online to photoshop and other tools but still there's crafts people who are really good at layout

And I think the same is true as we move from oil painting to digital tools and photography. I think the same will stay for that Assassin's Creed game you mentioned. Like there will be a lot of taste and a lot of art, even with AI acceleration. Do you think that AI can replicate that human taste and brilliance that it takes to create a great game? Like I was speaking with Demis. Let me put the context here. I was speaking with Demis Hassabis. Yeah.

who had his AI game player go and beat Go. AlphaGo beat LisaDahl, the grandmaster there. And I was talking to him about like, where do you want AI to go? And he says, I want to be able to not just play Go, but to prompt Go. And Go is obviously a game that has stood the test of time over the years.

And what Demis was saying is like, I am trying to build AI that can actually go ahead and build a game as brilliant and as elegant as those that designed Go eons ago. So let me put that to you. I want to ask you, do you think AI can do this? I think there's about five really interesting threads right now. I think we're still in the early forms of AI where arguably AI is

is super good at word manipulation and it's learning shape rotation and shape manipulation and more spatial type things and engineering things. So I think we're early there. 10,000 years from today, will AI be able to generate games and experiences that are maybe indistinguishable from humans?

I could see that. What's interesting, though, is thinking through what does that really mean for us culturally in a society between here and that next 10,000 years? Because there is a notion that

there's an optimistic notion that we still like things built by humans in a way as well. And we sometimes like to buy that designer thing rather than that mechanical thing. So I don't know how we go from here over the next 10,000 years. I do think ultimately AI is going to get pretty smart. Yeah, I would be...

At this point, I would say I'd be surprised if the AI can prompt a brilliant game like an Assassin's Creed in the long term. I imagine it will. So let me ask you a question about how you're doing this. You're building a 3D foundational model. Everything we hear about building foundational models is it's exceptionally expensive, takes the best AI talent to really do it well.

And it's only available to a few companies. Now, Roblox is not a small company. The day we're speaking, the market cap is north of $37 billion. But I want to know how you're able to do it with the resources that you have. Basically, whether you're able to build what you want to build and how tough this has been. It's super tough.

I was in a meeting yesterday. I join our AI group very often to check in on, I would say, not just our 3D foundational model, but our text gen and ultimately our experience generational model,

And we should also talk about, ultimately, NPC models as well, because that's kind of the sci-fi complement to 3D experience generation is avatar generation as well. We have a really large AI team, and the beauty of...

uh it's pretty big it's like um because thing if you look at our history we've we've made some acquisitions with loom ai and some brilliant talent there we've been constantly hiring and we've been building models for four years now primarily for text and voice safety for asset moderation we've

We've gotten really good at building and running AI at Roblox. Behind the scenes, there's over 200 different models running on our system that are doing so many things at higher quality and higher performance. We have also gotten very good at running relatively complex models efficiently. And DeepSeeks in the news, that's kind of...

rings the bell for us because our voice model runs very efficiently. We're able to use it to help moderate all the voice on our platform. And we've even open sourced one of our voice models so other people can use this model for safety and civility. So bridging that into 3D generation, we have a lot of experience with that.

We also have an amazing amount of data and a very...

Big data set that I mentioned is usable for training object, 3D object generation and started to put that together. So we're training. We have enough hardware. We have a great team. And we're going to run this efficiently so that any creator can use it in their game. So we think we're uniquely poised to generate this 3D foundational model and expand on that. And it's all proprietary? All proprietary. Okay.

Okay. And you're training on actual gameplay to be able to create more gameplay or not yet?

So this is very perceptive because we have both static information, which would be 3D objects and shapes. We have something more than static information, which is the code embedded in those objects. So cars on Roblox, we call them 4D cars, not 3D cars.

Because they have code, like when do you open the door? How does the user interface work on that? What are the properties of the wheels and the motors on that car? So we have that data to train on as well. How does a car actually function as opposed to how does it look good? The future, though, once again, in a privacy-compliant and IP-compliant way,

is we also will know more and more how people interact with objects. What's a typical way for you or I to walk through a building? How do people interact in a sword fight when they're pretending they're knights? How do people climb on cars? So that time-based information complements the static information and starts to beckon more interactive type experiences.

And I would say when I say proprietary, proprietary plus, of course, the huge open source ecosystem that's out there for tools, open source complementary things. But all of the 3D stuff, very heavily proprietary. OK, so what I'm hearing is you're training on the objects, some of the interactions, but not actual gameplay to be used to generate more gameplay. Not yet.

I would say not yet. And I would say training on gameplay isn't necessarily something we would use to create gameplay. But training on gameplay may allow when you or I pick your favorite historical figure, whether it's, you know, George Washington or, you know, the founder of any country, you

That data may allow us to make more natural avatar simulations as well. It may allow your scoopers to act more natural. If we know how those various character archetypes work when they've got a scoop and they want to share it with you. So I think we'll see that ultimately drive avatar simulation as well.

Okay. And now, David, I want to ask you, you're probably better positioned than most to answer this question, which is we've been asking, when can you prompt a blueprint, for instance? When can you prompt a design of a piece of furniture? The systems hallucinate too much right now to actually be able to do that in a useful way. At least those that I've seen, maybe someone's built one.

Having built this or being someone that's working on one of these foundational models for 3D, you're somebody that understands the spatial dynamics of something, how things interact with each other in the real world. Do you think that this type of technology can then be applied to be able to successfully be able to prompt a blueprint project?

prompt furniture design, prompt some sort of town layout, whatever permutation of this you might imagine. This is really interesting because there's probably hundreds of thousands of town layouts on Roblox. And there are probably, as we start to use the data of human motion through those town layouts,

There's, I'm sure, a lot of embedded information. What layouts work? What layouts promote better social cohesion? What town layouts aren't very good? And the combination of those existing town layouts plus how people interact with them does beckon a future of

Really the interesting thought of getting useful type of layouts from what's traditionally thought of as a game engine that could actually be used for real life purposes. I'll give you one example I just saw a few days ago. We asked our 3D foundational model to generate an over-the-road 18-wheel semi-trailer that

And we started constraining it a bit. We started saying, make it fit in this size. And these models are getting smart enough that in a smaller size model,

We'll generate the cab of the trailer without the trailer. And in a larger box, we'll say, oh, now you have enough space to generate the trailer. So I like the idea of a creator on Roblox who's thinking of it as a game using our AI to build a fun game called Build My Own House, have some AI help power it,

and have it be a mix of human intuition and AI to make the house really functional. And over in AutoCAD land or Replit land are these beautiful 3D architectural programs. Imagining those being driven by AI physics and aesthetics

I think we're all gonna be designing our own homes, absolutely. Okay, we've got some questions from our Discord, a lot of them about responsibility, starting with the AI responsibility question. We have a member that says, AI likely will soon be building its own worlds inside Roblox. Who actually takes responsibility if something goes wrong? Is it gonna be the user? Is it gonna be Roblox? Who's responsible if AI builds something bad?

Well, I think what we're going to see, wonderful question. And we are put in a position where we have an amazing amount of responsibility here of a wide range of ages. We're running in different countries with different regulations and policies. And where I see this going is we need to be able to handle this

whatever comes out of AI. And what that means is if one of our creators makes a really interesting 3D experience and they're running in a fairly untethered mode, we'll give them the guidelines on the type of experience, the age range, but it's our job to be watching that still. And I think it's our job to be using AI to

to analyze the output of what their AI built to see is it policy compliant. And so we take that responsibility. We take the responsibility of making sure you can predict and trust the policy adherence of what is on our platform.

Okay, so it will be Roblox's responsibility, not the creator that prompts. It is still the creator's responsibility. Like we have terms of service, do not build this. We do not tolerate this. We have zero tolerance for a wide range of things, including bullying, hate, you name it. But if someone breaks that policy, we're going to catch it and we're going to handle it. And so it's really both.

Okay, so no new precedent on digital liability if someone uses your prompting tool. No, I don't think so. What if the prompting tool hallucinates some weird shit? Humans can already hallucinate weird shit.

Yeah, but anyway, when you use a prompting tool, you give up some control. But I hear what you're saying. We can and we do. All right. Let's take a break. And we're going to come back and we're going to talk a little bit more about child safety, covering the Hindenburg Research Report. And then I want to talk about the metaverse. And I want to talk about financial responsibility within Roblox, which is something you and I spoke about offline. And I definitely want to get to. All right. We'll be back right after this.

Your ads deserve better. Smarter targeting, premium cross-channel placements, and simplified measurement all in one platform. StackAdapt is the leading ad buying platform for end-to-end performance. StackAdapt is ranked the number one DSP on G2. It's the ad platform you need for campaign success. See why the best marketers are switching to StackAdapt at go.stackadapt.com backslash LinkedIn.

I'm Tomer Cohen, LinkedIn's Chief Product Officer. If you're just as curious as I am about the way things are built, the insights behind what it takes to create a world-renowned product, then tune in to my podcast, Building One. There's so much to learn, like how Patagonia innovates with its supply chain. We had to go out to farmers and convince them...

It was really damn hard. Or the way Adobe thinks about the first interaction somebody has with Photoshop. I was always so fascinated by how people navigate and find their way. Ever wanted to know how Nike builds emotion into the Jordan brand? You have to be obsessed with the current state of the human condition. And it doesn't stop there. What about how Gleam reinvented knowledge search with AI? You can learn about how a Michelin star chef is redesigning seeds for flavor and how Pixar is nurturing a creative culture.

Listen to Building One on Apple Podcasts or wherever you get your podcasts. And we're back here on Big Technology Podcast with Roblox CEO Dave Bazzucchi. Dave, thank you so much for coming and speaking with us. I've been telling you, I told you right beforehand that I've been following Roblox for a very long time.

I've got younger cousins, friends, kids that use it, and I've gotten a chance to see the platform. And it's a fascinating place, and they do love it. So let's ask a couple of these safety questions. Just my job, I have to ask you them. So don't get too mad at me, but I've got to run through some of these accusations. Yeah.

We actually like talking about this and we feel it's a wonderful area for the future to really lean in on. So we don't, if anything, we welcome the discussion.

Okay, great. That's great to hear. So let me bring up some of the points that Hindenburg Research made and toss them over to you. First, they say that Roblox has lied to investors, regulators, and advertisers about the number of people on the platform, inflating a key metric by 25% to 42%. It's interesting that Hindenburg is no longer in business. Now, I'm not going to say there's any correlation between

Right. They're out of business now. Yes. But it is interesting they're no longer in business. And on this point in particular, you know, on TV publicly on earnings call, we completely disavow the whole allegation. And the interesting thing is investors can look at the ratio of the time on Roblox to

to the hours on Roblox, to the bookings on Roblox, that's available for four or five years. And you can watch the ratios of all of those. And when you dig into that math, it becomes very apparent that their research was completely flawed.

Now, they say that there's a potential that bots can inflate the numbers and there are some games that users have said are are bought it out. So do you think there's a chance that bots might have been increasing those numbers?

I would say, once again, we report daily active accounts on the platform. We are constantly, constantly patrolling for, we don't call them bots. We call them accelerated play or, you know, things that are doing weird thing. And the situation you're talking about, which no one on the platform wants is,

is some experiences, there can be an incentive to put in account farming type activity and resell that kind of thing. We're constantly getting better at that. We are constantly working on that. The numbers are really not significant. And once again, there's some interesting thing. If we imagined three quarters of the accounts on Roblox were bots,

Interesting. That would mean a quarter of those accounts are really making a lot of money per hour because the one thing that always pencils out is raw cash in our bank account.

That is audited. We go through that with auditors. We can measure it. And there's no way to bot raw cash. So, yeah, we disavow the significance of bots in really distorting our numbers. And that is the cash that Roblox players are spending within the game.

That is real cash spent by... So bots don't typically spend cash. And so we do think because our ratio relatively of cash intake to hours to DAUs is inspectable for four or five years, it highlights those kind of the...

The reason we push so back hard on those allegations. Now, one of the things they say is that you talked about how you track active accounts, but there could be one user with multiple accounts. And one of the things they say is that Roblox doesn't de-alt or basically bring down the number of players to individual users. It's more on the account level.

Yeah, I'll share this as a common practice. And it's one of the reasons why many companies report daily active accounts or daily active user accounts. It's because...

We wouldn't reject. You or I might have two Gmail accounts and we might use them for different things. So we count accounts, we count them every day, and we think it's the best way to report on what's happening on our platform. Right. Okay. Let's go over to Child Safety. Okay.

again, we'll mention Hindenburg is not in business, but they did say that their research revealed that Roblox is an X-rated pedophile hellscape, exposing children to grooming, pornography, violent content, and extremely abusive speech. Now, even if it's just a much smaller percentage than they talked about, that would be an issue. So how do you react to that? Yeah, I would go even further and say we are disturbed by

by even one bad issue on the platform. Like it's serious stuff for us. All of our employees, many of them have kids on the platform. I have cousins on the platform right now playing all the time. But every day, tens and tens and tens of millions of people are having an amazing experience on our platform. Right now, one issue is too much.

We constantly are improving and getting better. We shipped over 30 safety improvements in the last year. We had a huge release in November around parental controls and giving parents more control. And we run every single asset, every single image through AI today.

And then one thing people don't quite always realize is unlike most other platforms, on Roblox, there is no sharing of images. There is no posting of images. That does happen on a lot of other platforms. On Roblox, people are connecting and communicating. So we take it very seriously, and it's really something we really focus on. Right. They said that they registered as a child and found...

games like Escape to Epstein Island and Diddy Party, 600 Diddy games, including Survive Diddy and the Run from Diddy Simulator. I would say in pop culture, it's no excuse for us. Terms that sometimes are very welcome are

in five hours in our culture can all of a sudden turn into very bad terms. And I would say we're constantly accelerating the velocity of picking up things like that. Last thing on this. I mean, yeah, okay.

Last thing on this, they say that you reported a 2% year-over-year decline in trust and safety expenses as you were pushing to profitability in the second quarter of 2024. It would seem to me like as your users increase, that would be an investment that you'd want to increase. Yeah, what's interesting is they don't have access to our internal graphs anymore.

that we track so carefully. And I track all the time, we track in our staff meeting on the quality of our safety systems, on the accuracy of our safety systems,

on our vision really to protect people of all ages, not just under 13, but 13 through 17 and even 18 and up. And all of those systems are getting so much better continuously over time. The one thing they are noticing is

is in many areas we're accelerating AI and in many cases we're able to move people to more high-level roles in the company. So that's really an apple to oranges thing. We would push back and say all of our systems are getting better as we move to more AI on those systems.

Okay. And I was going to move on, but I do want to follow up on one thing you said, which is that culturally things can appear normal one day and the next thing you know, like allegations come out and they're different. But something like escape to Epstein Island or like survive Diddy, like there's no way that they were fine at one moment and then not in the other. I would say in those areas where

We have historically put an enormous focus on the actual 3D content. We are more and more incorporating the 3D experience as well as the description of the experience in a holistic way. And so we're always improving that. What does that mean? Well, I think what you're saying, what I'm saying is the 3D experience itself, which we've been very, very careful of in what you can actually do in that environment,

I'm agreeing with you. The title of experience can also matter and needs to be more responsive to cultural events. Right. Okay. All right. So I actually want to speak with you about something concrete that also came up in our Discord, which is that Apple recently said it's going to let parents share the kids ages to limit app access.

I'm very curious what you think about this move. Do you support it? And yeah, the question that we got here is, should that be expanded to more operating systems and codified into law? So I want to put that to you. This is a really good one. And we've taken the posture that we can't wait for mobile providers or OSs.

And we have to be fully capable of doing it on our own, whether it's in age estimation, how we filter all text, how we filter all voice, all of that. As these systems come to play and as we get a more reliable signal, maybe from a mobile phone,

We, of course, will use to the degree we can trust that signal as part of how we build our safety systems. But I would say we can't be waiting for that. And we have taken the posture we can't wait for that. But do you think other operating systems should follow Apple's lead here? Basically, it's a parent inputting the age of a child sort of to make sure that an app has that information and can gate content appropriately. Yeah.

I feel if we could see on a mobile phone or some other device that a person we can really trust as a parent has really said that is an eight-year-old, we would use that signal. And that signal would be helpful to us. But as far as...

whether between operating systems and our legislature, that's going to be a forced issue or voluntary issue. We're kind of building irrespective of it. Okay. All right. I want to wrap up talking about one of my favorite, uh, conversations here, uh, that I have with my friends, uh, because they have their kids. I don't have kids yet. Hopefully someday we will, uh,

And we're going to run into this issue, I'm sure, because they're going to be on their phones and iPads without a doubt as kids of a tech reporter. This comes with the territory. They beg the parents for money to spend on Roblox. And my friends are like, what do I do here? Do I give my daughter money to spend on the Internet and, you know, buy a dress on Roblox? Or do I like...

not give them that money or give them a physical dollar and not allow them to spend it online. So I'm very curious to put that question to you, CEO of Roblox. What would your response be? Well, first off, I remember when I was in kindergarten and I got 10 or 15 cents to go buy a comic book and that was my allowance. And every week I took that 15 cents and I walked down to the store and bought a comic book and

built some financial literacy. So that would be point number one. Point number two, the vast majority of people on Roblox are not using any money. And part of the reason we've grown is we've always viewed the platform as a free utility that brings people together with optimism and civility, no matter where they are in the world, and supports them. And we behind the scenes are

when we're discussing things like search and discovery and what are our algorithms and all of that stuff that's a lot in the news, behind the scenes, we are generally optimizing user joy, user retention, connection rather than money. And I think that's actually contributed to our growth as we put that first and foremost.

That said, yes, we build the company on the small percentage of

of people on our platform who use virtual currency, and that's going to be supplemented with other forms of monetization. And we're very appreciative of that because it allows us to build this platform and keep moving. When it comes to kids, I do feel it's 100% a parental choice. I would never recommend to a parent, "Oh, you should spend money on Roblox." I would recommend to a parent

to think through their situation, their family, financial literacy. And I would say the Roblox angle is I think more and more young people as they build financial literacy may balance physical world, financial with digital world. Like we already do that, right? We have subscriptions and we have online things with physical things.

And so I do think as far as if for those parents who are interested in financial literacy, this could be a kind of a compliment to that. There's another thing about it in that there in a way there are situations in Roblox where young people might be able to spend virtual currency online.

in ways that we would, you know, buying a ticket for an airplane or something that in a little bit of a way is a simulator of things to see later in life. So, you know, I would leave it to parents, but I think it's getting to be an interesting mix. Yeah, my advice to my friend is do it. Give them money. It doesn't have to be a lot. It'd be maybe a couple bucks a week or a month. And it is an amazing educational tool, I think, because it shows the kid that

You have a budget. You immediately have a budget if it's a dollar or two. And okay, like, is what they're buying the most valuable stuff in the world? Like a new shirt or a ski hat on their avatar? I don't know what you'll say. I'll say no. But I think that like, they're going to eventually...

end up in the real world and they're going to have, you know, a thousand dollars in their bank account and they're going to have to learn how to allocate that and then hopefully more. And I think that early budgeting is really helpful. I'm going to I would definitely do it. It's an interesting thing that someone will probably write a research study on. Is it better for kids to give one lump of virtual currency every month or

or to give a little bit of virtual currency every day, the once a month may have more educational financial literacy impact than the once a day thing.

Yeah. Like here's your budget. You want to spend it? Go ahead. But you're not getting any more until, you know, the next quote unquote paycheck. I mean, maybe they have to work for it. Maybe do some chores and you'll get two dollars in your Roblox account and you spend that on the hat or the skateboard. I don't know. David, I got to tell you, my avatar is ugly. I mean, I'm embarrassed to be walking around with mine. Yeah.

But I do think that like, yeah, it it's this thing where you learn work translates to money. Money is finite. Right. And this is what it does. If only all parents could copy your speech right there. This is your allowance. You're not getting any more. I thought that that was a good model for every parent in the world. Great. And so let's just close with this.

CEO of Roblox, you have 85 million users. What do the next five years look like for you if things go really well?

I'm super optimistic. Like we've said, we believe 10% of the world's gaming market can run on Roblox. That's our target. We think it's going to usher in a lot of optimism, a lot of human connection around the world, allowing more people to learn and create. And I do think also an opportunity to usher in

the use of 3D AI in a really safe and civil way. So I've never been more optimistic.

all right david look thank you so much for spending the time talking with me i learned a ton about where generative ai is going we talk so much about text generation voice generation 3d generation game generation a model that understands the world to me is the next frontier you're living that so that's that's quite cool maybe we get scoop hunter on the platform one day i don't know i want to play scoop hunter i'll send you it's live as a no as a cloud right now and

And yeah, definitely thank you for taking some of the tougher questions and for showing us where the platform is going. So great speaking with you and really hope to have you back sometime soon. Thank you, Alex. Awesome, David. Thank you so much. Thank you to the Roblox team for putting David on the show with me. And Ranjan and I will be back on Friday to break down the week's news. So we'll see you next time on Big Technology Podcast.