Company: Modl.ai
Headcount: 31
Stage: Series A
Tell us your story
My lifelong relationship with video games started with the Amstrad CPC 464 (I was born in its launch year!) which I inherited from my older brother a few years after its release. The trusty Amstrad introduced me to the basics of input/output operations, and how to load a computer game from a tape, making sure to get the tape counter value right. Come 1990, my relationship with video games was further cemented by the release of the Game Boy. What sealed the deal was getting access to “Klik & Play” as a ten year old: a game creation software suite that eventually turned into “Multimedia Fusion”. This convinced me that building video games was one of the most interesting things you could possibly do.
I sent my first invoice to a customer when I was around 14 years old. That sounds like I had a lot of entrepreneurial spirit, but the truth is I sent the invoice to a friend. I was interested in learning about contemporary computer kits but didn’t have the pocket money to fund my interest. To remedy this problem, I managed to get registered as a computer hardware vendor with a wholesaler, ordered components from them, and built gaming systems for my friends at wholesale prices. I was inadvertently undercutting the local computer shops (!) since I was doing it just to try out the hardware before I delivered it to the customer, not caring much about the margins.
When I graduated from high school (in Danish: gymnasium), I had convinced myself that video games did not hold prospects for a proper career and I should probably do something else. I turned to studying psychology instead. Psychology as it happens held ample opportunities for exploring human behavior and I soon found myself working on cognitive experimental psychology, psychometrics and simulations. All of these are key components of video game development I’d argue. I left to do a Masters in Video Game Development and a new understanding that it was in fact possible to work on video games as a career. I still worked as a consultant in psychometrics on the side working with the Danish Defense on high-impact personnel selection (think fighter pilots, special forces). I started thinking quite deeply about how humans make decisions in challenging situations and the possibility of using data to model or predict human behavior.
In 2011, I co-founded the independent game development studio Die Gute Fabrik alongside my friends Nils Deneken and Douglas Wilson. Nils had been running the studio since 2007 and invited Douglas and me to grow it into a proper business, and I operated the business as Managing Director for almost a decade. We launched several award-winning titles, including most notable Sportsfriends and Mutazione, the latter of which was brought to market under the auspices of the company’s fantastic current CEO, Hannah Nicklin. We even managed to win the Game Developers Choice Award for Innovation in 2012 ahead of several AAA titles.
It was during this time that I found my way back to academia by way of a project focused on treating PTSD with video games. I ended up at the IT University of Copenhagen, researching video games and artificial intelligence (AI). The work at the university quickly turned into a full-fledged enrollment into their Ph.D. program, through which I got to know Georgios N. Yannakakis, Sebastian Risi, and Julian Togelius, all later co-founders of modl.ai. My Ph.D. dissertation focused on using generative models for the automated testing of video games, the primary value proposition that modl.ai offers to game developers today. However, we didn’t start the company immediately; before we went commercial, I worked at New York University as a postdoctoral researcher and at Northeastern University as a tenure-track assistant professor.
In 2018 it became clear to Benedikte, Lars, Julian, Georgios, Sebastian, and me that both artificial intelligence as a technical field and the video games industry would soon collide, creating a pivotal moment. AI was undergoing a revolution in terms of methods and performance, which at the time was exemplified by solutions such as AlphaGo, AlphaZero, OpenAI five, and other examples of AI moving into new territory. At the same time, the game industry was displaying increasing openness to using modular solutions and to modern AI such as machine learning (ML) approaches. However, the intersection of the two opportunities was usually limited to the largest publishers and game studios.
We believed that these technologies could have value to the industry in general, which is why we started modl.ai: to provide an AI Engine to game developers who did not have the resources to build one on their own. This vision was much like Unity and Unreal Engine, which started providing game engines to the industry a decade earlier. Once we raised seed funding for the company, I left academia to support the company full-time, and ever since, I’ve worked at modl.ai as the company’s CEO, coming up to almost five years.
Tell us a story that has really resonated with you
The modern version of Westworld was a recent piece of popular culture. I liked it generally but it also struck a chord regarding modl.ai’s mission. At the surface level, Westworld is a dystopian tale about the potential consequences of sentient AI trained on human data and augmented with human manual design. However, if you go deeper, I think the story is fundamentally about what happens if we build technology that can observe, absorb, learn, and predict human behavior. Spoilers ahead for any fans of the show!
In Westworld, individual AI robots, “hosts,” have been trained on human data to simulate humans and are guided by narratives developed by other humans and eventually developed by other AIs. This is what most people think about when they think about Westworld. The show also features an imposing AI system called Rehoboam, which is the other side of the AI coin in the universe. Whereas the “hosts,” such as the pro/antagonist Dolores, simulate individuals, Rehoboam predicts the actions of everyone in society, working from prior data and behavioral modeling.
However, the data set needed for either application – individual simulation or general prediction – is really roughly the same. I think a mostly overlooked quality of Westworld was its ability to hypothesize what the availability of massive amounts of behavioral data encoded in AI could mean for human civilization in the far future.
At modl.ai, we work to collect behavioral data, model it, and leverage it to replicate human-like behavior. Our purpose is to provide artificial game players that can help a game’s success by playing it for testing and balancing purposes – or even play against actual human players to make the game onboarding or their long-term experience more interesting. Still, Westworld serves as a popular prediction of where generative behavior models could eventually take us and provides information about risks and opportunities inherent to the endeavor. There’s a hidden story in the show about imitating and predicting human behavior using AI that is obscured by the notion of the AIs becoming sentient. I think the show’s creators deserve more credit for this than they received!
What can't you stop thinking about?
Modularity in the games industry, associated fields, and, more broadly, how game engines are eating and will eat the world.
I think we’ve only seen the beginning of a trend. Writ large, this is the foundational idea of the Metaverse concept, but the vision has gotten a bit obscured by hype and confusion. I still think Matthew Ball and, of course, Neal Stephenson are holders of coherent visions of what one Metaverse could be. Much more pragmatically, the games industry will be the industry that provides the building blocks for any kind of Metaverse in the future if it comes to pass. And the industry has been rapidly modularizing for decades.
When I was just getting into game development in the mid-2000s, game engines, physics engines, and so forth were mostly things you would code yourself from scratch or leverage basic libraries for when you wanted to make a game. Now, almost 20 years later, you download fully fledged game engines from the internet and augment them with supporting products and modules depending on your use case. I think this trend will only continue in the years to come; the path to success in the video games industry and beyond follows the notion of modularity and the recombination of existing systems and functions.
In a world where the cost of video game development has been rising by a factor of 10 per decade, there is no other sustainable way forward than to modularize, reuse, and recombine technologies. I think we’re seeing early examples of this with generative AI applied to images, video, and audio now. Of course, our bet in modl.ai is that generative models for behavior will have an even greater impact.
If I could tell you just one thing...
Don’t worry. This, too, shall pass. If you live as authentically as you can with as much integrity as you can muster, you will probably be better off than if you try to live up to external benchmarks or expectations. Ultimately, I think it’s the best we can hope to do in life.
A little space for shout-outs
There are more people I would like to thank than could possibly fit here, so I would like to mention one specific person.
When I was in my early twenties, working in psychometrics, I spent a lot of time working with a psychologist and researcher at the Danish Defence College, Svend Erik Olsen. I spoke to him about my interest in pursuing research and, eventually, about my search for a good topic for a doctoral dissertation. Svend Erik suggested I look into modelling human decision-making. This became the bedrock idea of my work for the following decades and a core component of what we do at modl.ai. I think I owe him great thanks for setting me on that path.
If you could get a warm intro to anyone in the world, who would it be and why?
Well, I guess it’s far too late to ask for an introduction to David Bowie! I think artists often see what technology will do to society and vice versa, in the long term, before the rest of us who are merely building it.
This interview always comes to my mind when discussing the impact of technology on society. Sticking with musicians, I would enjoy an introduction to Wayne Coyne. I love the Flaming Lips and everything (I believe) their music stands for.