This week we’re bringing you a conversation with Hilary Mason from Hidden Door, an AI-driven narrative game engine. This mini-episode recorded in person at TechCrunch Disrupt and Dom and Hilary get into how generative AI is changing online gaming, building a team of creatives, and fundraising in the gaming space.
This week we’re bringing you a conversation with Hilary Mason from Hidden Door, an AI-driven narrative game engine. This mini-episode recorded in person at TechCrunch Disrupt and Dom and Hilary get into how generative AI is changing online gaming, building a team of creatives, and fundraising in the gaming space.
Found posts every Friday. Subscribe on Apple, Spotify or wherever you listen to podcasts to be alerted when new episodes drop. Check out the other TechCrunch podcasts: Equity, and Chain Reaction.
Connect with us:
Dom Davis 0:02
Hey everyone, it's Dominic Midori Davis and you're listening to found the podcast where we give you the story behind the startups. This week, I'm flying solo and bringing you a special episode I recorded at disrupt, I got the chance to talk with Hilary Mason from hidden door about the future of AI and gaming. So here's that conversation. Hi, everyone, this is Dominic Midori Davis, one of the CO hosts of bounce and we are live, as you can hear from TechCrunch Disrupt in San Francisco. I'm here today with Hilary Mason from hidden door. How are you?
Hilary Mason 0:35
I'm great. How are you doing today?
Dom Davis 0:37
I'm doing good. You know, it's not bad. SF weather's Sunny. It's gorgeous. Talk to me about hidden door.
Hilary Mason 0:44
So what we do at hidden door is take any work of fiction, like a book or a TV show or a movie. And we make it a playable social role playing game automatically. So for players, if you're the kind of person who reads a book, and you keep thinking about it, or you watch the show, and you want to yell at the TV, this is a way for you to come into those worlds, and be able to direct your own stories in those worlds automatically. So our players make their characters, they can play with their friends and family. And they get to direct the kinds of stories they want to play. And it is a game that plays very much like a tabletop role playing game with those sorts of dynamics. And it comes together as a dynamic webcomic that you see, as you make choices that's generated just for you add on the content side for our authors and our other content partners. They're able to take the book they wrote or the show they made and create a new experience for their fans where their fans can get really immersed in the worlds that they've fallen in love with and get excited for what might be met. So we're finding a lot of excitement there.
Dom Davis 1:48
Yeah, you know, I saw that, like Netflix had something where they had like some TV show, and it was like you can have your own experience in it. And you know what I'm talking about, like interactive and stuff. And I remember thinking like, how do you do that? Like, how do you build out a world like this where you can direct the story?
Hilary Mason 2:02
Well, I think you're referring to Bandersnatch, which was a really interesting experiment where they actually gave you a bunch of pads, you could follow and you could choose what would happen next. But even that was pre written content. So there was a team there who created all of the potential endings ahead of time. And then the players or the viewers sort of got to pick how they wanted to navigate through and how they wanted to experience it, it's really cool. What we do at hidden door is a little bit different, because there is no pre written contents. The whole thing is something that can be directed by the player, we also are a little bit different in that we've invested heavily in the idea of a controllable story engine. And so our authors are able to set rules that are consistent and how their worlds are experienced by their players. So what happens at the end is a story that's a collaboration between the original author the system and what the system is capable of, and then the player and their choices. And it's completely unique. And there's no pre scripted content at all. Yeah, that's
Dom Davis 3:05
so cool. So how do you choose the stories that you let people play with,
Hilary Mason 3:09
so we partnered directly with the writers, so we licensed their work, we make sure they're getting paid, and we make sure they approve of everything we do.
Dom Davis 3:17
Oh, my goodness, is there like a dream story would lower a dream author you would love to work with?
Hilary Mason 3:22
There are so many, I am a big sci fi nerd. And so one that I keep coming back to is altered carbon by Richard Morgan that got made into a TV show, I think was a Netflix show, but I'm not sure this year. But mainly, I think what's really fun about these stories is that you're either able to like sort of explore and poke at new aspects of a world. So you can go to places you haven't seen before. And any world in which it's not really about a drama happening in one character's head, but it's a world that you want to explore works really well. Other ones that I'd love to see are bridgerton, for example, like where there's some sort of relationship based drama and a lot of status games to play. Yeah, I could keep going. But there are a lot of fun things out there.
Dom Davis 4:05
I was gonna ask which genres Lin best to it or like, is there like a Pride and Prejudice one? Or is it a laundry
Hilary Mason 4:10
list? We haven't done it yet. But yes, there will be. And I would say that broadly, we sort of think of it as sort of like action based plots, exploration based plots, and then what I'll call drama, but that's it could be politics, it could be sort of social climbing, it could be creating and like becoming the mayor of something, or it could be falling in love is sort of those dramatic kinds of plots. And all of those things work well. What doesn't work well are things that require deep detailed factual knowledge of the real world or stories that take place entirely inside someone's head. So those are things we're not as excited about.
Dom Davis 4:49
And can you talk to me about the AI component within hidden door?
Hilary Mason 4:53
Yeah, I mean, this is only possible because of the AI system and I think it gives us to real afford and says, what's really exciting for our players is that they can have a game that plays like a tabletop role playing game. Where to do that right now you generally need a person to invest a whole bunch of time in creating and guiding you through a story, you need to buy a bunch of books, learn a bunch of rules, you probably need to spend at least one I played a lot, it's been like four hour sessions around a table. And it's incredibly fun. But something that is such a big time and energy commitment that it's really not as accessible. And so using our technology gives you that play experience, but in a way where you can pick it up and play for 30 minutes with your friends, nobody has to do the homework, you can put it down again. And then on the content side, it allows us to take say a novel, and create that world to plan in a matter of hours or days. Whereas game development typically would take 12 to 18 months around a property without the underlying tech.
Dom Davis 5:59
What has it been like fundraising this idea? Like how do you how do you break this down to investors? Yeah, it's
Hilary Mason 6:04
been a really interesting fundraising journey. And I want to say we've been incredibly fortunate and the folks we've gotten to work with, but it definitely has changed from three and a half, nearly four years ago, when we started this company. And GPT two came out while we were thinking about it, and we'd come in, and people would be like, you know, you're doing what, or last year, when we raised our seed, it was definitely all about like, Oh, are you putting NF T's in this because you could raise more. And so we've had to come in and explain what the underlying capability was, and then explain the game on top of it, and the business model. And now everybody knows the capability is interesting. And people have sort of caught on to the idea that what we now call generative AI is a set of fairly transformational capabilities. So it's almost gone too far. The other way, we're now there's a tremendous amount of hype around it. But I don't think the hype has caught up with a pragmatic common sense. And so now we get a lot of inbound, but I think it's actually quite shallow, which I probably shouldn't say, but But it's, you
Dom Davis 7:04
know, Yeah, cuz basically, you've you've been pitching a product that had AI, before, it was like trendy to investors. So it's like three or four years ago, what were investors saying, were they interested in the AI component,
Hilary Mason 7:15
I'd say it was a mix, like, we obviously got very lucky and found people who understood both the technology and the vision for the players that it opens up for us. But we had to do a lot of work coming into the room to explain what the tech was, and what was possible and what it was good for. And what we were actually willing to claim was true and isn't true. And so for example, we sent in our very first pitch deck, our system is not creative, our authors are creative, our players are creative, the system is essentially a combinatorial troupe machine, that those people can direct to create a really fun experience together. And I think coming to it, apart from the hype, sort of let us bring some people along, but I will say there were definitely rejections we got because people were just like, you know, you're you're a few weird nerds. And, and I don't want to spend the time on it, or I don't really see the business or we actually got one great rejection, which was if you do this, no one will ever stop playing it and you're gonna destroy society. So I can't invest, which was my favorite rejection ever. I don't know if it was serious or not. But it's pretty funny to us. Oh, they got the landscape has changed a lot. And between our preceded now we also had this whole sort of divergence into the world of blockchain and games and blockchain and AI. And that's not what we're doing. So we sort of had to navigate that as well as something a little bit different.
Dom Davis 8:33
Yeah. How would you even put an NFT? Or how would you even use NFT? Because imagine if you would have did what that investor said. And now NF T's are like not cool anymore.
Hilary Mason 8:41
Well, I mean, I'm a technologist, and very much a pragmatist. And so, you know, when you do play a game, you do create unique items and characters and new locations and new things that have never really existed in the world before. And they're cool, and you can share them. But they're, you know, honestly, they're each aligned in our Postgres database, and we have one of them. And so folks would look at it and be like, Oh, that could be an NF. T. And, you know, we just say like, Okay, why, you know, if we reach a point where that's what our players demand, sure, but we're definitely not seeing that utility got. And I'm very glad we didn't go down that path and sort of stayed programmatic. And I wanted to talk a
Dom Davis 9:18
little bit about what you've been before hidden door, because you have a very sciency background, you have this really extensive career. Why did you decide to pivot into gaming and AI?
Hilary Mason 9:27
Yeah, I have to admit, I've been a gamer my whole life. So this was a passion already. But you're right. I've been working in machine learning and data science for 20 years if you count my detour as being a computer science professor, and my last company was called Fast Forward labs and we did applied machine learning, research and product co development largely with Fortune 500 partners, so very much an enterprise company. Our customers paid us for access to our research and then sometimes we build stuff with them as well. But we did work in natural language generation and 20 14, we did work in summarization with deep learning in 2015 and abstractive summarization. And I could not get essentially how much fun the things you could build with this were out of my head. And so when my co founder and I were, you know, we knew we wanted to work together, we had an inkling that it would be in this technical space. But we really started playing around with many, honestly, many different product ideas. And the one that just seemed to be both the most interesting like, basically was the thing that we wanted, like, the thing we couldn't get out of our heads was to create this kind of gaming experience. And the rest of it came after that. And we also started thinking about safety and controllability. And we realized that standard LLM architectures began at the time it was GPT, two leading up to GPT. Three would not give us those things. And so we built a very different architecture, we do some LLM under the hood, but we built it around that controllability, which now gives us the ability to work with content partners, and give them confidence that their worlds are going to be what they want it to be. So it came together piece by piece. But it was just that obsession with play, and the obsession with the tech. And we don't have
Dom Davis 11:07
much time left, but I really wanted to get to this. So how big is your company? Now?
Hilary Mason 11:11
We are 14 people,
Dom Davis 11:13
14 people, how would you describe yourself as a leader?
Hilary Mason 11:18
I think pragmatic, direct and hopefully focused and maybe a little bit chaotic. My job as a leader is to build the machine that builds our product and business. And we've pulled together a team of folks from gaming and entertainment from machine learning. And somehow three and a half years, then I've gotten us all to work together. And so I think it it really comes down to that.
Dom Davis 11:39
Do you have any tips on how other founders can kind of foster positive workplace environments and keep everyone energized to work
Hilary Mason 11:48
for you. And I think it's having a lot of clarity over communicating, especially if you're working remote. And just being a good human being does a lot of it.
Dom Davis 11:57
What would you say is one challenge that you've experienced while building and scaling this company that taught you a lot about being a founder and an entrepreneur,
Hilary Mason 12:06
I have many challenges, I can pick a ton of them. The one that I'll probably mention for this audience is just that when you're a founder, your job is everything and it's nothing. So your job is to focus on what's most important, make sure everyone knows what that is, and then try and build the team such that you don't have to do that work anymore. And it means that you have to give up a lot. So like there are areas of our product that I built myself, like in our very first prototype, and I've had to take my hands off and step away and let people who are frankly, able to think about it all the time and better at it than I am come in and carry it forward. I think that one's been one that's hard for me to learn,
Dom Davis 12:44
and actually was something I wanted to ask you regarding the AI element. How do you train your algorithm? And cuz I'm wondering how does the bias and AI conversation kind of intersect with your company? Yeah,
Hilary Mason 12:55
that's a? That's a great question. So again, we started around this idea of safety and controllability. And that actually did come from, I think, a fairly robust background and set of values around building ethical systems that don't discriminate. And MLMs do magnify the bias in the underlying data. And they are not controllable, like by nature. And so I don't want to go into our whole infrastructure. But we essentially have pre generated have human eyes on most of the pre generated content. And then we use additional machine learning systems, which are, by the way, a mix of language models, and then other things like hunger and models to be able to contextualize and color it and create what the player sees. So in essence, we have players input and text we structured in our game engine, we use that to generate what comes out the other side. So we have a lot of control. And that allows us to control for bias among other things. And that's really, really important when you build products with this talk. And
Dom Davis 13:51
I actually have two more questions really quick because I'm really curious. So you only work with like authors and books or can you also adapt like, I don't know Broadway musicals anything? Okay.
Hilary Mason 14:00
We've never done a musical though who?
Dom Davis 14:02
Okay, great
Hilary Mason 14:02
idea.
Dom Davis 14:03
That'll be really fun. Last question, what's next?
Hilary Mason 14:07
So we are launching in November and if anyone listening wants to play, go to hidden door.co And then we are really focusing on building an amazing player experience from there and you will see us work with a whole bunch of different kinds of content over the next year and see what people love.
Dom Davis 14:23
That's so exciting. Thank you for stopping by.
Hilary Mason 14:25
Thank you.
Becca Szkutak 14:28
Found is hosted by myself. TechCrunch Senior Reporter Becca Skuta. Alongside Senior Reporter Dominique Midori Davis founders produced by Maggie Stamets with editing by Cal our Illustrator is Bryce Durbin funds audience development and social media is managed by Morgan Liddell Alisa stringer and Natalie Kreisman TechCrunch is audio products are managed by Henry pick of it. Thanks for listening, and we'll be back next week.
Transcribed by https://otter.ai