Will Wright Interview
Feed magazine just published a really cool interview with Will Wright. Get a peek into his vision of the future of The Sims!
One unavoidable thought confronts you after spending time with Will Wright at his offices in Walnut Creek, California: this man has the greatest job in the world. Cocooned in a small interior room in the Maxis headquarters fifteen minutes east of Oakland, Wright works at a desk covered with tinkertoy-like objects, which he absent-mindedly assembles and disassembles while talking. As the creator of some of the most original and commercially successful media properties of our time -- starting with 1990's SimCity, and culminating in last year's blockbuster, The Sims -- Wright has plenty to talk about, roaming effortlessly from the architectural musings of Chris Alexander to adaptive software to the surprisingly thorny problem of keeping the Sims from running into each other.
For the hour or so that I spent with him at Maxis, the only thing that he seemed to enjoy more than talking about his creations was actually playing with them. Halfway through our conversation, he logged onto an early internal test of the online version of the Sims and was immediately pulled into the game, engaged in a kind of Survivor-like popularity contest with other members of the Maxis team. From the sneak preview that Wright offered, the game looks to be even more addictive than his previous titles -- a merger of The Sims and SimCity with a networked component that has always been strangely missing in the Maxis simulators. Wright's switching between broad theorizing and furious game-playing speaks to exactly what makes his work so extraordinary: his best titles manage to be both intellectually rich and fiendishly entertaining at the same time. When I left his office, he was still hunched over the monitor, building alliances and creating worlds.
-- Steven Johnson
FEED: Let's start with The Sims. I know you've gotten this question a lot, but why no weekends in the game?
WILL WRIGHT: That's actually a very long explanation. Basically, we didn't want to try and do a scheduling game, and we didn't want every day to be the same. We wanted every day to be as different as possible. We didn't want to put the user in the position of having to wait four days before they could take time off to do X, Y, or Z. There are so many things in the game where all of a sudden you'll just have to take a day off -- you have a baby, say. So we said instead: "You can always take a day off. But you can never take two days off in a row." Also, if we'd included weekends, we would have had to show the days of the week going by. At that point, you'd start wanting a calendar. It just seems natural to be showing a calendar, which then makes it easier to start thinking about things like seasons, which we didn't have in the game.
FEED: It's funny, but in creating both Sim City and The Sims, you got to the idea kind of obliquely. You started doing one thing -- you're building a background cities for a helicopter game, or an architecture simulator -- and then you ended up doing something else. Is that just the way that you work, getting distracted by things?
WRIGHT: Well, I shouldn't call it distraction -- it's more like serendipity. I'm usually attracted to a subject and I get really lost in the subject and I start working on it. And then, I'll start trying the thing from the beginning in terms of how can I present this subject to somebody else. And then, on the way to that, I'll see core avenues, and slight deviations, and then paths that branch off. But even with The Sims, it started basically as an architecture game, but I think it got most of the architecture stuff, I wanted into the final product. You had to populate the world to evaluate the success of the architecture, but it wasn't like the house-design element dropped out altogether.
FEED: How did you develop the spoken language of the characters in The Sims?
WRIGHT: Most people don't realize how much we actually recorded for that -- we actually recorded hundreds and hundreds of different emotional nuances. And so, in fact, you are getting a pretty good reading on their emotional state during the conversation. You just have no idea what the topic is. And the little balloons above their heads give you kind of like some clue as to the interest matching between the characters, whether they're getting along or not. But it was really amazing compared to actually having scripted dialogue where you instantly would say, "Oh, it's a script."
FEED: Playing the game, you don't mind the repetition as much, because it's not a script, you know? I mean, most games come to the point where you're like, "Oh, man, if I hear that character say that one line again, I'm going to just jump out the window."
WRIGHT: [laughs] Right. In fact we did a lot of experimentation to find a language that would be the hardest to discern repetition in. And, in fact, we were originally going to do real languages, and we experimented with like three or four of them and they were all too repeatable. So eventually the voice actors ended up inventing this language. Actually, this one guy invented it and had to train another voice actor, and then train another one. And they'd have these long conversations back and forth, kind of teaching each other the language. It was fascinating.
FEED: And can they still speak it? Are they out doing conferences now?
WRIGHT: We started out with Navajo, actually, which is a nice language.
FEED: Navajo's the one that they used for that old encryption scheme, right?
WRIGHT: Right, and still very few people speak it. But we couldn't find any professional voice actors that speak Navajo. So we went to Estonian -- Estonian is actually really cool, too, because it's very hard to place. It doesn't sound European, it doesn't sound Asian. Doesn't sound New World. But we only found one voice actor who spoke Estonian, but we couldn't find a female, so we ended up going with that made-up language.
FEED: We're clearly still a few steps away from games where spontaneous behavior really starts to happen and you have real learning going on with the virtual characters. But were there cases with The Sims where you pressed play and the characters really surprised you with their behavior?
WRIGHT: A lot of stuff, yeah. Probably the strangest thing was how smart they were. I mean -- overly so, in a problematic way. We came up with this algorithm where they could kind of sense their environment; they could sense their internal state and they could decide, "Okay, what would maximize my happiness the most right now?" And in fact, the algorithm ended up being way too good. So in fact, we had to dumb them down quite a bit: We made them rather irrational and we had the personality values figure in much higher in their decision-making process than it did at first.
FEED: How do you think we can get to the point where our virtual characters generate spontaneous behavior on their own -- you know, where you look in one day and your character has suddenly learned how to juggle. Is that like decades away? Or is spontaneous group behavior more promising?
WRIGHT: With groups forming higher-level behaviors -- that's only going to occur when there is some feedback of the success of the group behavior back to the individual level. Without that feedback, you'll never see that happen. There are ways to build those feedbacks into the system right now. Frequently in most adaptive systems, though, the real computational bottleneck is the fitness test -- not generating new behavior. It's determining how good an existing behavior is. And that's what takes the most amount of time. I mean, that's a process of life, you know -- how long does it take to measure the fitness of a person? Maybe seventy years. You know, route planning might be one of the first places we see this kind of learning. People don't realize what a total pain in the ass route planning is in a game.
FEED: What does that mean, really?
WRIGHT: Just how to get a character from point A to point B without getting stuck behind a door or bumping into other characters. What makes route planning really hard isn't so much the static case but when you have other objects moving in the environment, they have to react to them. That's what makes it really hard. So like in The Sims, when they're walking from one room to another and other Sims are getting in their way, they might have to turn around and they kind of get cornered -- they actually have to calculate how quickly they can turn that angle. Then they actually calculate the angle of displacement from step to step. Most people don't realize how complex this stuff is -- in this example, you can't have them in a full stride and all of a sudden turn ninety degrees. Whereas if they're walking very slowly they can easily turn ninety degrees. So there's a relationship between velocity and angle of displacement and it gets rather complicated.
FEED: It's amazing how rarely the Sims get stuck -- I noticed that when I first started playing the game.
WRIGHT: Yeah, we spent a lot of time getting that right.
FEED: And that's an area where you should be able to kind of grow some intelligence?
WRIGHT: Yeah, because that's something that's very domain specific. So frequently in The Sims, the way they get stuck right now -- currently, it's very similar situations over and over. They're specific to this game. And so if they could actually just learn what the situations are and avoid those, that'd be a vast improvement because, based on their local knowledge, they could build specific heuristics for this specific system. That would really be a much more efficient way of solving the problem than classic solutions -- which is to have a programmer sit there and try twenty different algorithms.
FEED: Tell me about what you're working on now?
WRIGHT: Well, the main project right now is a version of The Sims that's online. It's going to come out some time next year. It's going to be kind of a massive persistent world, you know, fifty to one hundred thousand people in a city -- kind of like Sim City from the top down except there'll be houses, real houses made by real people who are logged into the game. Basically, you log in -- you'll have three slots to create characters. When you create a new character, you'll kind of pick the head and the body, kind of like the way you do with regular Sims.
Then we're going to have what we're calling different cities. These are actually going to be really pretty large areas. Once you pick a city, you get stuck in an apartment to begin with. The first screens show you what neighborhood where you're living in, and then you start zooming in -- it shows you exactly what house. And then you get plopped into the house. At this point you're playing something like The Sims, except everybody in the game is real, not computer-generated Sims.
I can actually see stats about my house, about my city. There are a couple of things I can ask about the world. People can actually create neighborhoods. They can decide: I'm part of Camelot. If enough people say they're part of Camelot in the same vicinity, a neighborhood will be created, and there'll be a big sign there. There will be a whole message board with people who live in Camelot. They can chat and have group activities and stuff. I can show where are my friends and they'll show me my friends, where they are online in the city. One of the things we've got -- and this is one of the cool parts of this -- is that we've got this concept of friendship. There's going to be this popularity game that's actually played. If you have popular people that are your friends, your popularity goes up. And so one of the sub-games of the whole thing is going to be this popularity game that we'll play -- this whole backstabbing, high school thing. That's actually a very important concept of the game.
One of the things we're really trying to do with the game design is we're trying to drive people -- players -- to entertain each other. And so one of the reward structures in the game is going to be built around having you attract people to your house. You don't have to build a house. A lot of people are going to do like a comedy club or bar and grill or maybe a theme park. Whatever they want to build. And whatever it takes to entertain and attract other people, they get rewarded for. And so we're trying to build the whole game around the players being creative, to entertain the other players. Also in the game, I can do searches like a Web browser. I can search for specific people, houses, neighborhoods. I can also bookmark -- bookmark my friends, bookmark houses. So this is kind of what we're calling the matchmaker. Actually we're having a play session right now that I can show you -- we have a rough prototype. We've been playing this game for the last two weeks and we're playing this very cutthroat popularity game. And we've been keeping our identities secret. Right now, my character "Mother Theresa" is the top of the heap. I have the highest popularity -- but nobody knows it's me. So when I go to this game, they're going to come attack me. They're going to try and bring me down. It's actually been amazingly cutthroat. There's one person in particular I need to go after. I need to find Brittany. [Looking at the screen] Okay, here's someone -- let's go talk to him. I need to build up my popularity with this guy. [Pause while interacting with the game.] See, now we're talking. This is good. I wonder if he wants to kiss?
FEED: They seem like a really nice couple.
WRIGHT: Yeah, I know.
FEED: Except I didn't think Mother Theresa dated all that much. [The characters kiss oncreen, syrupy music swells from the computer speakers.]
WRIGHT: There we go!
FEED: Oh, that's disturbing.
WRIGHT: That's going to help me a lot. I need to basically beef up my popularity so I can go after Brittany. She's the only one who's close to me in winning this game.