Darryl Long can’t make his Ultima IV, his favorite game from his formative days. So he works on artificial intelligence and cloud computing to help other Ubisoft Entertainment designers and developers make their Ultima IV.
Long is the managing director of Ubisoft Winnipeg, which the French publisher opened last year. Winnipeg is a sort of headquarters for the company’s artificial intelligence and cloud computing efforts. These technologies help Ubisoft use its 32 studios across the world — as one of its comms reps told me, the sun never sets on Ubisoft — to work on massive game worlds such as Assassin’s Creed: Odyssey and live-service online games such as The Division II.
The AI helps free up animators and artists to do more than work on patches of grass and other mundane tasks. You can designate an area and set boundaries, and the AI will fill it in with grass and trees that look as if an artist had created it — and use foliage that fits the terrain and environmental conditions, even erosion. Or someone at Ubisoft Bucharest can work on a scene, and another designer in Montreal can look at the material as the Romanian studio continues to produce it.
I found this fascinating, and Long walked me through a few of the basics behind Winnipeg’s work at a Game Developers Conference 2019 meeting. Here is an edited transcript of our interview.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
A far cry from before
GameBeat: You talked about procedural algorithms. Looking at the most recent games Ubisoft has put out, what can you point to where you used a procedural algorithm to do a certain thing?
Darryl Long: The best example I can give is Far Cry 5, because I produced Far Cry 5. I just came over recently. There’s a gentleman named Etienne Carrier. He did a presentation at GDC last year, specifically on this topic. The biggest one I could point you at is the vegetation in our worlds. Back on previous Far Cry games, we had a system to place trees and bushes and all that in the world, and then we added the ability to place rocks and have that be decoration for the forest. The artists needed to draw the boundaries of that region by hand. It was a lot of work. What we did is we automated that using procedural algorithms. We would take the terrain of the world and simulate how much light it gets, how much moisture it gets, what the temperature gradients are like, things like that, and find out what kinds of trees would grow there, what kinds of bushes and grass and whatever. And then figure out how much they would grow, even, in that location. Something with a lot of wind, the trees don’t grow very high. The vegetation in Far Cry 5, the vast majority of it was build through a procedural algorithm following those rules.
GamesBeat: So for example, here’s a meadow, and the algorithm would know that the grass would grow so high and the trees would be this tall depending on how the wind comes down the mountain to the west.
Long: Absolutely. Even further than that, it would decide where the meadow should go sometimes.
GamesBeat: Based on the terrain of the area?
Long: It could be that. It could be — on the sides of these mountains in Montana, the grasslands tend to run down the peak of the erosion along the side of the mountain. Then the trees grow more in the valleys in between those. It’s rules that are programmed into the procedural algorithm, and then it knows these things and it generates terrain. It’s strong enough that it can keep them fresh and make sure it doesn’t feel repetitive.
GamesBeat: Would this also apply to Far Cry: New Dawn.
Long: I didn’t work on New Dawn, but we would absolutely — at Ubisoft, we want to make sure we’re not reinventing the wheel all the time. Of course we want to pass technology on from one game to the next.
GamesBeat: Say 10 years ago, you’d have a team of artists making this by hand. Not even talking about doing the boundaries for the algorithm, but making it all by hand.
Long: Back on a game like — I’m trying to go back far enough to remember what I was working on. It never got released, but there was a game that I worked on 15 or 16 years ago. When we make grass, we’re not actually placing every single blade of grass specifically, but we do grass cards, is what we call them. The grass cards were all placed by hand. An artist would say, I’ll put a grass card here, a grass card there, and do a meadow that way. It took an incredible amount of time, and of course that limits the size of the world we can do. Nobody’s going to want to create 16 meadows every day.
Above: This church has gone to the dogs.
GamesBeat: And it limits what else those artists can be doing.
Long: Absolutely. I’d much rather that an artist was building a new mission or something like that than placing blades of grass. It’s about empowering our creators. If an artist can click a button and have grass everywhere, and then decide how they want to shape that grass and give a certain feel to it afterward, that’s a tool that adds value. That enhances the game they can make. If we can create tools that allow them to say, “I want a house,” and they click a button and get a house, already built and filled with beds and tables and televisions and all that stuff, then they can customize it. They can go in and tell a story. That’s something I believe artists can do, but that machines will never be able to do: to tell a story. Put the soul into a creation that makes it really special and memorable for the player. When you walk into an environment, into a house or a building or something that’s been crafted in a game, you immediately and unconsciously pick up on the story being told in that scene. I don’t believe it’s machines that can do that for us. It’s artists.
GamesBeat: That also gives you the time and the resources to talk to these geologists and meteorologists and herbalists to tell you how these things behave in nature.
Long: For me, that was one of the great things about Far Cry 5. Doing that kind of research allowed us to connect with people who understand how the world works and reflect that in the game. We had people come to us and say: I grew up in Montana, and this feels like home.
GamesBeat: The people that use these procedural algorithms on your teams, they could be from any of the creative disciplines that are using that engine? It could be someone working on Mario vs. Rabbids, doing something kind of cutesy. But do The Division folks say, we like that, and they can grab that through the cloud?
Long: That’s the idea behind this technology. All of the cloud becomes a platform on which we build our games, and that allows us to share seamlessly between the different teams around the world. If you generate an algorithm, or build an algorithm that can generate meadows, that’s something another team could take and add meadows to their game. If they change the assets that go into it — the type of grass, say — they could have a totally new environment compared to what was originally created.
Above: Name a more iconic duo. Go ahead, I’ll wait.
GamesBeat: Could the Mario team use an algorithm the Far Cry team created to make that kind of cartoony grass?
Long: Yeah, absolutely. It shouldn’t be any problem at all.
In the cloud
GamesBeat: Are you on something like Azure or Amazon, or does Ubisoft use its own cloud resources?
Long: We’re actually working with many different cloud providers. We’re on a lot of them.
GamesBeat: Is it a case that you just need so much cloud that you go to different providers?
Long: Generally, no, that’s not really the limitation we’re seeing right now. It’s more about how distributed their infrastructure is, how far they’ve spread.
GamesBeat: And that’s important, because as a member of your comms team quipped out there, the sun never sets on Ubisoft.
Long: [Laughs] It’s true. On Far Cry 5 we were working with the team in Shanghai. They’re 12 hours offset from Montreal. We were working on the game around the clock.
GamesBeat: You used motion capture as an example of another thing there. You smooth that out with an algorithm. How many people would that take in the past?
Long: To do the motion capture cleanup? It’s more a matter of how many seconds of animation you can clean up per day. I don’t remember anymore.
GamesBeat: So it’s more about, if you have an algorithm to do this, you can do more mocap, and that’s where you get characters like Cassandra from Assassin’s Creed: Odyssey.
Long: There were times where we would capture a scene, and we wouldn’t get it back for a couple of weeks because of the processing of the data. For us as developers, of course we want it right away. We want to get it in the game and start iterating with it. Reducing that from a couple of weeks to a day or less is a huge win for us. That’s the kind of efficiency we’re getting.
GamesBeat: Winnipeg doesn’t make games as much as it helps everyone at Ubisoft make games.
Long: Actually, we are working directly with the teams. We’re building solutions that are custom tailored to particular teams, and then taking that and sharing that with them first, because that’s what Ubisoft does. But we’re working directly with the game teams.
GamesBeat: Does Winnipeg work with every game team at Ubisoft?
Long: No. We’re working with several of the major brands already, but yeah, there’s a lot of teams at Ubisoft.
GamesBeat: Is the ambition for Winnipeg to eventually take these solutions and apply them across the company to the point where there’s a Winnipeg person working with every team?
Long:To me, the ambition is that we create a platform to build our games that people can use, that the game teams around Ubisoft can use if it suits what they’re doing and it makes sense for the project they’re making. We’d definitely be happy to work with them all.
GamesBeat: Using the cloud, the folks at Blue Byte could work on something, turn it in at the end of the day in Germany. At the beginning of the day in Montreal, those guys could start work on it?
Long: I’d go much further than that. Our ambition is that we’d be able to be working on the same data at the same time. We want to collaborate across the world. To me, the ultimate is that all of our developers across the world can be working on — sorry, I’ll phrase that differently. Any of our developers across the world can be working on the same game at the same time.
GamesBeat: Are you worried about crunch with that? For example, say the Blue Byte team has something they really want to work on, but they know the people they want to work with at Montreal aren’t around until 8 p.m. their time.
Long: It’s a bit of human nature, and I know game developers love the work they do. They want it to be the best it can be. I think the teams absolutely are happy to share at Ubisoft, and they’re going to be glad there are other people collaborating and contributing to what they’re doing.
GamesBeat: But you’re not worried that this possibility of having everything available any time across the company could lead to people working hours that are too long?
Long: I believe that the teams will embrace this and see it as a solution to make their jobs better, and allow them to make better games.
GamesBeat: I know that having a 24-hour cycle to work all day helps, but has anyone ever talked about having a unified work schedule at Ubisoft now: Let’s have people work the same eight hours across the world, and we’ll just call it Ubi time?
Long: [Laughs] It sounds like something from a science fiction movie. So no, no Ubi time.
Team matters
GamesBeat: When you take a look at your AI team, is it just one team in one place making these algorithms and working on procedural generation? Or is it groups of teams with different studios?
Long: There’s an initiative in our community of studios called La Forge. They have a pod in Montreal and a pod in Toronto. That team is doing a lot of machine learning work, working with local universities. Our Winnipeg studio, though, our studio as a whole is using machine learning to develop these new tools to help us build better worlds. That’s ongoing right now. Are you asking if we’re sending people to other studios?
GamesBeat: No, I’m asking if the AI team is concentrated in one or two studios, or if it’s distributed across the company.
Long: Many of the teams across Ubisoft — it’s a company of 15,000 people, right? Many of the teams are looking at how they can use AI is to improve the way they build their games. Ours is the first studio dedicated to improving the way we build worlds.
Above: Ubisoft Mumbai works with the Indian Institute of Technology Mumbai & Intuit Lab, one of India’s foremost universities.
GamesBeat: I know that a lot of studios Ubisoft opens up, there’s always a partnership with a local college. Is there a partnership like that at Winnipeg?
Long: We’re talking with quite a few local universities. We’re a new studio. We’ve been operating for six months, basically. We’re talking to the University of Manitoba, the University of Winnipeg, MITT, Red River College, just to name a few. We’re also talking to some of the high schools. Community engagement is important for us. We’re putting down roots in this city.
GamesBeat: How many people are at Winnipeg?
Long: Right now it’s funny. The answer changes every time I ask, because we’re growing so quickly. Right now we’re at 35 people.
GamesBeat: Would you ever want to make a game led by Winnipeg?
Long: Would I ever? Realistically, we’re not going to be — right now our commitment is to go to 100 people in five years. Building a triple-A game, building a big Assassin’s Creed or Far Cry or anything like that, takes a huge amount of human power. That’s something, right now, that’s not on our road map. It’s not what we’re thinking about. We’re thinking about how we can help the other Ubisoft teams build their games better, how we can learn in Winnipeg. This is a city that’s now just creating its video game industry.
GamesBeat: It’s not a hub like Edmonton or Montreal.
Long: Exactly. There are some smaller studios in Winnipeg already, but we’re the first big triple-A publisher in Winnipeg. We need to learn. We need to be humble. We need to learn from the other studios and get their knowledge first. Right now this is all a learning phase, seeing how we can contribute to other games.
GamesBeat: Your Ultima IV has to wait.
Long: Ultima IV will have to wait. [Laughs]
GamesBeat: You cited Ultima IV as one of your guiding games. Looking at today’s landscape, though, what’s a game that means a lot to you?
Long: There’s a ton of different games I’ve played. What’s going on in the industry right now is inspirational. Sometimes it’s a little frightening to see how fast things are moving. It’s great.
GamesBeat: I went to an interesting talk by Jason Rorher, who does a bunch of indie games, about the “Indie Apocalypse.” What he learned was that it was the shorter, contained experiences that weren’t doing so well, but games like Rust, games he called “unique situation generators,” were doing very well. I look at that and I see a parallel with the online live service model, games like For Honor and The Division. Is that something Ubisoft has figured out, how to make these games live for years and grow over those years?
Long: Rainbow Six is a great example of a game that’s been very successful and lasted a long time. It’s had three years of live operations now, I believe. I know that one of—for us, it’s important that we create experiences that players can engage in and invest in emotionally. Once a player has done that, of course they want to stay with that experience as long as possible. I think that is a trend in the industry right now. People want to find a game, fall in love with it, and stay with it for a long time.
GamesBeat: What’s a good example of a machine learning concept or algorithm that you use with these live service games. Is there something that, say, Rainbow Six couldn’t happen without?
Long: I can’t give any specific examples for Rainbow Six. I’m not informed enough on what they use to build that.
