“Were you the one that said I was a super-villain?” It’s not the most auspicious question to be asked during an interview. But when I first saw Justin Lyon speak, at the GaME14 conference at London’s Imperial College, I thought he had a lot of the ‘90s corporate supervillain about him. Something of the Jonathan Pryce in Tomorrow Never Dies. After all, he’s handsome, waspishly-smart and wears the standard-issue accoutrements of a corporate overlord. And his company is called Simudyne, which sounds like Cyberdyne. And they mainly work for governments and huge corporations. And their slogan is ‘Engineering reality’. Even their logo looks evil.
So, Simudyne has the accoutrements of a 1980s villain company. But do they do evil? Well, no. They create simulations. Do people do evil in their simulations? Well, no, not intentionally. Looking at their carefully-anonymised case studies, you can see that they do it for all sorts of people, many of whom are either working for good, in government or in big business. (There may be some crossover there.) From a quick scan, I can see simulations covering disaster management for the port authorities of the Western US, training banks in counter-terrorism, managing deepwater oil drilling, and a recreation of one of Microsoft’s headquarters office blocks for simulating physical and cyber attacks.
The reason that they’re making these simulations is because the heads of corporations and governments get things wrong, badly. “They’re serious men -- usually men.” says Lyon. “They’re serious men doing serious things and they don’t have time for these rigorous games. Yet you look at the spreadsheets and the tools that they’re using that the consultants built for them, and it’s a joke mathematically. Lots of detail complexity, but the mathematics literally, and I say this repeatedly, literally the ancient Egyptians would understand it if we could get past the language barrier. It’s an illusion of understanding.”
Yet when Lyon was playing Civilization he’d feel like the game was helping him gradually grasp the whole of a complex system, rather than simplifying it. It also helped him learn about managing it. “You would compress thousands of years into two or three days of intensive play. Then you restarted and you would get better and better. Now, I had meetings with very senior politicians, and you’d go into the mayor’s office, or any major corporation’s, or even to a Prime Minister’s. You look at the tools they’re using and they don’t have the ability to experiment safely. They don’t have a Civilization for Great Britain. They don’t have a SimCity for Chicago. You say “Why not?” This was built in 1991, surely we can do better than red, amber, green power points in the 21st century.” And that’s Simudyne’s aim.
To create these simulations, Lyon was inspired by Jay Forrester’s System Dynamics, the same field that inspired Will Wright to make Sim City, but throwing in a soupçon of complexity-science and agent-based modelling. “Whereas Will Wright was inspired to use this mathematics for safe cities to build a game, I was inspired to do exactly the same thing, but for real cities.”
To make these simulations, Simudyne employs a small development team who rapidly iterate a model, then create a UI for the client to explore it. “We use a combination, I suppose, of mathematics to capture the physics of how the real world works and then put it in pictures, 3D like The Sims, or 2.5D like Sim City or Farmville or maybe even like 2D like Pokémon or Donkey Kong. They’re not games, but they look a lot like games.”
And that’s why we’re talking to Lyon, because the things his company makes look very game-y, though he balks at the phrase ‘serious games’. What Simudyne makes are simulations of real world situations that have appropriated the design and UI of games to make them intelligible to modern executives. Not because these executives are gamers, but because game UI is so far ahead of UI in any other field, and executives are not always the most technical people. “In the early days of Simudyne, our models and our simulations were very difficult if you were not trained. We would struggle because the insights were absolutely profound, yet because they looked like spaghetti diagrams to executives, they were unable to connect with the insights.”
So Simudyne started building simulations that were explicitly game-like, with four specialised employee types to make systems, pull in real-world data and then make it accessible. “Engineers are the humans that in conversation with the client, listen and then translate that into diagrams, mathematics. The programmers then quantify that into the code on our Simucore system on the cloud, so that the computer can understand it. The designers make a pretty picture so that the humans cannot get bumped down by all the math. The integrators are just grabbing the data from all different data sources and pulling that into the simulation.” These teams consist of between 3 and 15 people. In a couple of months, much like an indie developer, they build a simulation using off-the-shelf tech and assets that the business can use to test and make important decisions.
One of the simulations they’ve made was a simulation of how a port operates They linked up several of these simulations, got 200 port authority managers, coastguard and homeland security guys into a hotel, and put different teams in charge of different ports in different rooms. Then they destroyed one of the ports in an earthquake, and watched all the different teams struggle with the changed situation. (I want to play this simulation just to see if I can be Frank Sobotka.)
Using design principles learned from games has allowed Lyon’s teams to draw non-gamers into their simulations, so that they're more willing to engage with them. “By having the beautiful user interface we can draw the humans into experimenting safely. The more you play a video game the more likely you are ‘win it’ or to navigate it more successfully if there’s no sense of winning. Correct? Similarly with the simulations we build, the more times the human plays it the more knowledgeable they’re going to be about a situation. What we found is in the past, if the simulation was boring or it was too mathematically heavy, then humans would only play it one, two or three times. They didn’t get the rich learning that we wanted them to get.”
Drawing them in is key because so many of the strategies and tactics thrown up by the engineers’ highly complex mathematical systems are counter-intuitive. “Now, (the executives) say ‘Well, no. The math must be wrong, the model must be wrong. My mental model is accurate.’ That’s because their mental model is based on flawed functions, because they’re limited by human cognitive biases... Humans over millions of years have evolved incredible pattern recognition skills. What we are unable to do is solve non-linear equations in our head. No one can. I don’t care how clever you think you are. You can’t do it. It’s only in this collaboration with our computers that we can overcome it.”
The obvious follow-up question to that, from anyone who’s grown up on a diet of Asimov and Arthur C. Clarke, is why we still need people at all, beyond setting the criteria for a simulation. “It’s not computers versus the humans. It’s this robust and rich dialogue between the things that we’ve invented and the things that have evolved. It’s together that we can really solve some of the world’s most pressing problems, that’s what we believe.” Lyon thinks that humans in collaboration with computers make radically better decisions, and that either alone is inferior.
Though he only thinks the computers are inferior for now. “We can run tens of thousands of simulations and identify resilient and optimal strategy. It might take six months or a year for the humans to wrap their head around it and implement how to change management processes to actually execute. That’s unfortunate… at some point we won’t have the time to allow the humans to beat their head around what the right decision is. The computer will actually simulate, will actually execute.”
So his ultimate aim with Simudyne is for it eventually to be the spine to the data-centric world. He envisages natural-language processing programs like Siri to be the interface, that services like Wolfram Alpha will parse those queries into searches, and that Simudyne will run the simulations - “millions or trillions in the time it takes to snap your fingers” and provide the answers.
To do that, their simulations need to be interoperable, whereas the everyday sims we play are standalone, monolithic. This also means they have to be less perfect than you’d expect, only accurate where it matters. “The map is not the territory; our simulations are not trying to copy everything in the real world. It’s completely unnecessary and actually counter-productive. A good simulation tends to get simpler over time as you winnow away all of the extraneous aspects and focus in on the core physics. Once you’ve validated that then you start to connect it to other ones.”
He shows me a quick example video, of a simulation of a hospital they’ve made for a major American city. The simulation is used to test out evacuations in a variety of emergency scenarios - “They don’t practice that very often in the real world, if at all because it’s too dangerous. These are people that are on life support. They can’t move them to practice a drill.” - and then he shows me how this simulation is embedded in a wider simulation of the city, so you can see the fallout of that emergency across other systems and services. A power failure in the hospital cascades across all the other city’s embedded simulations.
So what’s Lyon ultimate aim? “Simudyne is about transforming how all decisions of consequence are made… an operating system for the world.” An operating system that ties together government and big business, and makes everyone think they’re playing games, to build a near-perfect sim, before ultimately handing control over to AIs.
Like I said - a supervillain. Albeit, a well-intentioned one.
Correction: An earlier version of this article stated the name of Simudyne's founder as Justin Lyons, when it should have been Justin Lyon.