The Quest to Make Massive Gaming Worlds Realistically Complex

You're in a massive online gaming world. You meet a character who says he needs you to collect ten coins. You complete the task, and the character thanks you. You find a monster and kill it, then run into an unmarked boundary; the landscape stretches on but you can't access it.

Another player enters the world. They come across the same character. He says he needs them to collect ten coins. Your contribution is forgotten. The monster is mysteriously revived and they hit the same invisible wall.

Although gaming worlds today can look extraordinarily complex, they're generally limited in terms of how much you can interact with their virtual content, and what impact your actions have. Impr​obab​le, a London-based startup, wants to change that.

Improbable is working to build massive simulated worlds; environments where you can interact with everything in real-time alongside many other players, and where your actions have repercussions for the rest of the game's explorers. See a bar stool? You can kick it over. And when the next person comes along, it won't have magically righted itself.

"The basic impulse behind the company was really simple: We were wondering why certain game experiences that we wanted in online worlds just consistently didn't exist," explained 27-year-old cofounder and CEO Herman Narula in the company's Farringdon office. "Little things, like being able to travel across the whole online world without any barriers or walls, or being able to drop something and then come back and it still be there."

It's the sort of experience he reckons non-gamers imagine games to be like. "They're meant to be worlds where you lose yourself," he said. "But they're not really."

Narula first had the idea for Improbable just before his final exams at Cambridge University in 2012, where he studied computer science. The company started out in a barn outside his house. Initially, he said, they had no idea of the technical complexity behind the issues he wanted to address.

Now, the team thinks they've cracked it. And they're not the only ones: Last month, US venture capital firm Andreessen Horowitz invested ​$20 million in the company.

While it started out as a frustrated gamer's dream, Improbable's solution is not just for games. It turns out that building large-scale dynamic simulated worlds can have many applications. Take a realistically complex simulation of a real-world city. You could look at how people use public transport or how disease propagates. Narula said that they're currently working with biologists, economists, and even the Ministry of Defence (he couldn't give me more details on that last one).

On the gaming side, they're working with London developers Bossa Studios on the upcoming massively multiplayer online game Worlds ​Adrift as well as New Zealand game designer Dean​ Hall, who was behind award-winning zombie game DayZ.

So how do they do it? Narula describes Improbable as an "operating system" that games and other applications can run on. The idea is to get around the limitations of running everything on a single game engine or server, or a set of engines each representing one part of a world and stitched together. Narula complained that this results in a world that's "completely like cardboard" with limited objects to interact with. "The way they create fun because they can't simulate the world properly is they create loads of simulated quests, and loads of scripted events," he said.

To get past this, Improbable is dispensing with the idea of having one job to one game server. The key, he explained, was to split up the work.

The company takes thousands of "workers"—which could be game engines but also other applications—and coordinates work across them with no central point of control, relying on distributed algorithms to keep everything going.

The workers wouldn't all be on one machine; they'd be passed around servers in real-time to be near the other workers they needed to interact with and reduce latency.

Narula conceded that the real technical details are "special sauce-y," but said they had mechanisms to effectively tell the workers where the work was at any one time and how to cluster, and how to replace workers that die—which would be inevitable given the numbers involved.

"Think of it as literally like a server cluster which has a dynamic topology, and it's passing hot potatoes around it in real time while all these workers are working, without that process of swapping them interrupting their work," Narula said. "That would be why this is really, really hard."

The whole thing runs on the cloud and the underlying operating system remains the same for every new game or application. Every time a developer builds something, that can be added to a library to be repurposed for other projects.​

Narula recognises how it sounds: Improbable is promising a lot, and people likely won't be convinced until they experience it for themselves. He compared it to trying to market an anti-ageing cream; users aren't really going to believe it until they see results for themselves. Worlds Adrift is likely to be the first opportunity for the public to try it out, sometime later this year.

Right now, Narula said their main constraint is getting enough people on board to meet growing demand. They expect things to get easier as they move forward; at the moment they're having to work very closely with developers to train them in the new system.

Where could it lead ultimately? There are many things that academics, industry, and other organisations want to model with a level of realism that correlates to reality. For instance, one project Improbable is currently working with intends to investigate the UK housing market by modelling entities such as properties, buyers, and sellers, and the dynamics between them under different conditions. Improbable's technology could let them do this on a much larger scale.

On the gaming side, I asked Narula if being able to simulate a more detailed environment could lead to a virtual world that's close to the real thing; a kind of hyper-realistic Second Life. He said that, potentially, it could—but while Improbable is tackling the computational restriction of building a world where so many entities can interact, you still have to build "all the shiny content" to go in it.

And the question then arises as to how realistic you might actually want it to be.

Narula suggested that being able to behave in an unpredictable way could make for a positive gaming experience. "These memorable events, like the kind of things you have when you go on a night out or have a really good holiday—those are all about incidental, accidental collisions and interactions with other people," he said. But is it still fun if a player, for instance, decides to burn everything to the ground?

The rules will be down to developers of individual games, but Narula reckons that you need some negative events to keep things exciting.

"When you read a novel, bad things sometimes happen to the character, and that's part of the reason why it's interesting," he said. "Worlds where I die and it means nothing, or where I can't ever lose—that can be very hollow."

Goodbye, Meatbags is a series on Motherboard about the waning relevance of the human physical form. Follow along here.