Is there an AI model of the dog in terms of its attributes, and the AI model in terms of like, building relationships or things like that? How does that work? AS: Yeah. It has a model of what its base personality is, and we're probably going to expose some of the players to that, some not exposed yet. Like, is the dog, how social is it? How greedy, or generous, or tolerant, or energetic -- all these various personality traits. And then those determine what is going to make another dog attracted to it. So some dogs are really attracted to high-status -- dogs that have accumulated lots of toys, for example. That might attract a sort of materialistic dog. And some other dogs might be sort of introverted, and want another dog that's also introverted. Or maybe some dogs are more conservative; so, they meet another dog who's got, like, six romances going on. Maybe at first they get wooed by them, but once they meet some other girlfriend or boyfriend of that dog, now they're totally turned off. So there's a little model. It's not super-sophisticated; we have to keep it simple. How does player agency fit into that? Some of these things are values that can reflect the way the player perceives, or wants to aim, their dog. AS: Yeah. So when they first adopt the dog, we're going to expose some of the personality. So you're already making a decision, in the adoption center -- do I want a dog that's promiscuous? [laughter] Do I want a dog[-friendly] dog? There'll be some decision there. And then you can influence your dog's personality, to some degree, depending on how you take care of it. You know, the identity, try to shape more of it with these missions, can be an influence. And during the play date, you can buy gifts in the pet store to give to your dog, to give to other dogs. Or depending how I dress my dog up, make him dress up in an embarrassing outfit or not, you know, that might influence things. If my dogs start to scuffle, and I decide -- do I pull them apart, or keep fighting? So there's various factors, and the idea that the Dog Feed is explaining to you, what is the reasoning behind what just happened? So, yeah, the player has agency there. But the dog also has a mind of his own, too. So it's like a collaboration between you and the dog. Are you working with anyone who has a background in social networking or web development or is that sort of like stuff that you just taken as an influence? AS: Yes, and its influence. We're learning as we go. We're working closely with these guys here [at ngmoco] of course. And we've been working on some stuff for a while now. But yeah, definitely surveying what's out there and learning what works and what doesn't work. Adding Social Networking: The Sharable Player Story It's interesting, and the fact -- you talked about it in the demo -- the Dog Feed is a status report on the interactions your dog has had. And I'm assuming it's going to cover the missions you talked about, where the dog goes off and you send him on missions and, also, the relationships he has with your friends' dogs. AS: Yeah. And, maybe, any other things that you just happen to do. If I train my dog to now catch a Frisbee, then that will be announced, too. Also, dogs that need to be fed or that no one has been playing with, it will announce that they need that. It's a whole huge mix of stuff. The key to making this interesting for a lot of people is having dog relationships between themselves and other players. Especially in San Francisco, almost everyone I know has an iPhone, and probably a bunch of people are going to play this game. But you still have to make that leap to both agree to buy the game and then network together. What kind of challenges do you see in terms of getting that sort of social networking aspect up and running in a game context? AS: Word of mouth is obviously huge with these kinds of games. We are trying to make this so that the game is so cool that I would want to show you. Right? "How cool is this? You have to get this game so that we can have our dogs play together!" We are thinking of other mechanisms for ways to have players find other players. There will be a global feed, too, where you can view activity going on in general over the whole network. You could potentially find other players who you've never met that way. The social networking aspect of it is something that I think is probably going to be compelling for people. Obviously, that's why you put it in there. People like to compare, and I think people like to compete in different ways. That is probably going to drive it. AS: Yeah. Some of it will be competition -- like "my stats are higher than yours." Or I see cool stuff that you're doing with your dog, and I'm going to want to do that. But I think some of it -- I don't know if "cooperation" is the word, but the "play" is going to be fun, to make our dogs go through relationships. Oh, our dogs were in love, now they've broken up and now they're back together again. That's great. KK: It's like a soap opera. AS: Yeah. There's some competition and there is some just play. KK: It's a story. AS: Yeah, a story. A sort of a procedural narrative emerges from the dog's life. AS: Yeah. Is there any sort of arc to it? Is it a chronology of events, or is there an arc that you see emerging? AS: As your dog's social life gets richer over time, I think the dog will eventually need to reflect that. If two dogs had been together in the past, they'd broken up and they got back together, then you have now progressed to that point. They are reunited. And with those missions, there is definitely a progression that you make in terms of achievements in the missions. There are some arcs there. In a way, with soap operas, you can go on forever. Right? The way that soap operas do, things happening again... It is always interesting to see why: "Why are we friends or not, today?" There are also moods. We can still be friends, but our dogs could be annoyed at each other today for some reason. Then they would recover from that. Or our dogs are in a romance, and then my dog is attracted to some other dog but hasn't started a romance with that other dog. I think it will be, hopefully, endlessly entertaining to see that unfold. Tell me about the design of the missions. You alluded to the missions, but we weren't able to check it out. What I got was that you pick a mission and send your dog on it. You don't see that mission happening but you get the result. AS: You'll get a screen shot, though, of what happened. You'll see your dog rescuing a cat off of Mount Everest. Then, on the Dog Feed, if you click on that item, Dog Feed, anyone can see that image. The State of A.I. Today What do you think of this state of the virtual pet genre? I'm not asking you specifically talk about any project, but do you think it's evolved, or do think there's a lot of room to evolve? AS: Yeah, there's a lot [of room]... it needs to evolve more. Yeah, we need to see more new stuff with the genre. It seems to me that with the current capacity of any of the platforms that I have mentioned, iPhone, DS, anything, there is probably more things to be done, but the simulation aspect is very weak. I would think probably in terms of behavioral simulation, there's room for improvement. AS: Yeah. That's the hardest thing to do. That's no surprise that most virtual pet products are a little light on the AI and the simulation part. It's hard. It takes a lot of expertise and a lot of time. So... What's the benefit to concentrating on it? What's the concrete benefit to concentrating on that part? AS: It's coolness. It's cool, the actual product, the quality in the product. Just shipping it and get help to sell products. Do you think that's what the end user is going to perceive that? KK: Even when I play it, after so many months, I'll see the puppy do something that's emergent or a little different that I've not seen before and it's really cool. You start to form a better connection with it. AS: You just believe in it as a living thing. It's not fake. It's not simple, or obvious... it feels more alive. That's the goal at least. That's what we are striving for. What about communicating that to the player? AS: There's going to be more of everything you see here [indicates screen with dog on it] but there's going to be even more facial expression. Like if he's hungry he's not going to be running around happily -- if he is starved, really needing to be fed. He's going to whine and howl and yip. With characters, there's sort of the underpinning of AI but what's important is the way the AI expresses itself. KK: Yeah. If it's not expressive it doesn't matter. It doesn't matter how much affect is going on if it is not showing it, right? It goes back to -- how does the simulation engage the player? AS: Right. Have you talked to people, like other professionals who are working on AI? Have you found that there is a demand for people wanting to push forward? This is going to sound accusatory; I thought that AI would push forward more in this generation than maybe it had, in terms of behavior. AS: Yeah. I think a lot of us want it to. I think the reality is that large studios... I know a lot of the AI programmers and designers, and they all of course want to push and innovate. But it has a lot of constraints. They have to ship product. They can't take too much technical risk. So, it's really the smaller studios and indie studios that can afford to take technical risks and innovate. I think that is where a lot of the innovation will come from, like ngmoco. KK: My total personal opinion, but I think the expansion of online gaming has actually slowed down a little bit with the perceived value, from the business side. Because when you have other humans to play against... AS: Yeah. That is true. With the rise with online gaming, AI perhaps takes a little bit more of a backseat for a little while. I've heard people say that explicitly, at least in terms of, "Who can make a better opponent than another person?" AS: Well, but it's not about opponents. It's about cool characters that are creating, potentially, a story. So, it's about when you need strong characters. The problem is that you can make small, incremental innovations in AI, working in a large studio. But to really make a revolutionary leap, you have to take more risks, and you have to take more time. It's tough to do within larger companies. It's part of how larger companies work, I think. There's that, combined with the fact that it is a really hard problem, even if you have a lot of time. I spent five years, off and on, working on the Façade stuff. It was like three person-years for my collaborator and me. We made some progress, but it is a long way to go.