itm2

Rod Humble talks Sims, A.I., and his new work at ToyTalk

In 2014, Rod Humble left his post as the CEO of Second Life developer Linden Lab. After releasing solo project Cults & Daggers and a techno album, he has returned to games to work with ToyTalk, a game company featuring Pixar vets that has so far specialized in interactive apps for children. Most interestingly, though ToyTalk stimulates interaction between children and artificial intelligence through actual, organic conversation. Humble spoke to Kill Screen founder Jamin Warren about The Sims, his plans to expand the company’s audience to more than just children, and the magic that occurs when we can talk to a machine. [—Savannah Tanbusch]

Kill Screen founder Jamin Warren (KS): How did you end up at Toytalk?

Rod Humble (RH): I wanted to take some time, like, in my mind, I had that I’d take a year off after I left Linden Lab. I just wanted to recharge my batteries, but that turned into—because I apparently can’t sit and do nothing—it turned into, well, I’ll release a techno album and I’ll make a game in a year. So I promised my wife that after a year, I’ll do a real job again; I spent a year and made a techno album and a game about Hellenistic-era religions, so I got that out of my system! Then after a year, ToyTalk had reached out and it’s a great opportunity. It’s just unusual enough to fit what I like to do so it was great.

KS: Were you familiar with ToyTalk ahead of time?

RH: Yeah, a little bit.

KS: Had you met ToyTalk founder Oren Jacob before?

RH: I was vaguely aware of ToyTalk but after they reached out I realised how big the ambitions were and to actually make this new artform of computer conversation was really interesting to me. The company had taken steps before with apps for a younger audience, but now the technology is ready to branch out into other areas. That was super exciting.

“do you really want to miss out on the opportunity to make a new kind of artform?” 

KS: Was there a particular “aha” moment in talking with Oren? How did you know you were basically sold?

RH: Yeah, actually, it was at one of the dinners. We were talking about the different kind of things that we could do and he said, “Do you really want to miss out on the opportunity to make a new kind of artform?” And I was like, “Well, that’s a good point.” I don’t want to miss out on that, that’s what I kind of do. I like that. At the risk of being a bore, I think that games as an artform and new kinds of interactive artforms are what I really love to get my teeth into. So, those were sort of the magic words when the lightbulb went on.

KS: You’ve been working in games and working on AI problems for quite some time. If you were to go back 20 years ago, do you feel like ToyTalk is solving or potentially approaching some issues? Is there a through-line between some of your early work?

RH: I do, I do. And you know, storytelling has been one of those holy grail things that a lot of people in games have looked at. “Can we make really good, memorable characters? Can we actually solve the conversation problem?” There’s a whole bunch of those issues. Certainly, working on The Sims we bumped into this a lot. It’s part of the reason Simlish exists, because you want this emotional expression, but language was kind of a bridge too far. And, I think really it’s been over the past 5 years that there’s been a bunch of breakthroughs in artificial intelligence speech recognition. The technology is now good enough. You have chatbots that pass the Turing test in one way or another. Chatbots that are now mainstream in terms of their use and people wanting to interact with them, and them being believable. And voice recognition being this other massive investment for future technology. People talk a lot about virtual reality, but voice technology has been a little bit more under the radar, despite the amount of money being poured into the space. Like any good group of artists, I think the focus of ToyTalk, and now myself, is that, you know, that we can subvert and steal that technology for the cause of art and entertainment. So it does address a lot of long-standing issues in games. You’ve looked at the history of conversation in games, as well. Do we have branching? Do we have the past stuff from text adventures? How is this going to work? Conversation with speech solves a lot of issues, but no way do we claim to have the complete solution yet. That’s kind of what we’re working on.

KS: So what will you be doing with ToyTalk, and how does it build on your past career working in games?

RH: We’ll be looking at new audiences, so I think previously the products have primarily been for a younger audience. Going forward, I think we’ll be looking for broader audiences. I’m prototyping a whole bunch of small projects, some that I’ve done in the past and seeing which interactive mechanics fit with with computer conversation. And, those have got to be new mechanics. Frankly, you can’t have a regular game with voice thrown on top and you can’t have a movie with some puzzles thrown in. A lot of the work I’m doing right now is on new mechanics. But you know, to try to answer your question, things like drama, those areas that are traditionally ripe for conversational media are of interest. The goal is: can we ever get as good as an interactive version of a Who’s on First sketch, or, for the more high-brow version, if we could make something as good as an interactive version of a Chekhov short story, then we win! So that’s the aim and we see how far we get before we fall flat on our face.

KS: We rarely think of voices as being a mechanic, right, like that being an option. I’m thinking about TellTale Games, for example. There, choice is a mechanic, but there’s a set number of responses, there’s a set number of outcomes. They’re working on a lot of different problems, but one of the things that’s not available is that I can’t speak my answer. Or even a game like Mass Effect, there’s lots of games that present choices, but it’s rare that I decide what the character is actually going to say. Is that a new problem in games? How do you encourage people to ask good questions? Because that’s something that, as players, we typically don’t do. It’s rare that the ball is kicked back in my court, and it’s like, “Oh, do something creative with this. Try to find a creative way around whatever it is you’re trying to do.”

RH: You put your finger right on one of the hardest problems. The aim is that we put you in a conversation with a character and naturally you will feel the freedom to go anywhere you want, but the setting and the situation would encourage you to take these certain paths. And so we have to cover both sides, and we have to cover the chatbot side of, well you know, what if you’re talking to a character and you decide well, I’m going to talk about the weather, instead of the fact that the submarine we’re in together is about to run out of air. You’re like “Oh, I wanna know what the weather is like upstairs.” We have to cover that, but at the same time, we have to cover in-depth the actual situation you’re in and make that feel believable. So we have to do both, and that really is new. We do have an open mic and we do have this open input so we have to broaden our consideration. Some of that legwork is handled by the artificial intelligence and the storytelling AI in the background and some of it has to be handcrafted, and so it’s this blended approach.

computer conversation lends itself to humour. 

KS: I always loved the bugs in The Sims. Whenever that comes out each year for The Sims, I always love reading through the sort of unique problems that a game like The Sims create. Has there been any surprising bugs or surprising problems that you’ve faced in the prototyping process that you sort of weren’t expecting?

RH: Yeah! Actually there is. First of all, I have been blown away how naturally computer conversation lends itself to humour. Some of the early prototypes—literally I programmed them in like a week—we do this Friday demo drink meeting, and the company’s like, just rolling. And these are people who have worked at Pixar, and so, they’ve actually got a decent bar on what comedy looks like. The reason is that when you talk to something that is clearly inappropriate—an inappropriate object. I won’t use the internal stuff that we’ve been doing, but an example is, have you ever seen Fawlty Towers?

KS: Yeah, it was on PBS when I was a kid.

RH: There was this one sketch where his car breaks down, and he’s literally having a conversation with his car and he ends up giving it a good thrashing. And it’s really funny, right, because we all talk to our car, but the spectacle of seeing someone do it is funny and, I can assure you, it’s even funnier when you do it in an interactive medium. And, so that surprised me: how natural comedy feels because for most game mechanics comedy doesn’t feel right, outside of slapstick. Slapstick feels right because we get a lot of physical slapstick games, like Goat Simulator and all of those. That’s funny because we get these physical reactions, which again are inappropriate for that setting and that’s kind of what makes it funny. I think the other one is how it’s really touching when you say to a piece of entertainment software, that is looking at you, in character, “I love you.” I just said it into the phone and I’m saying it to you, and you feel something, right? It’s like “Oh!” How many times in games have you ever said “I love you?” There’s a lot of emotion with that and, it actually, really ratchets it up. So, those things have been very surprising to me. Just the act of speaking changes the context.

KS: Yeah. That’s a great observation. What can you say about the demos that you’re working on?

RH: I don’t think I can go into specifics, but I can give you the broad context. One involves various planning mechanics, so for example, a planning mechanic is something you see everyday in sports games or soccer games. “Hey, when you get out there guys, you go on the left, you go on the right,” and then you watch it play out. So that is an area that is pretty ripe for conversation. It’s like, “Hey, when you get into the building, look out. If they ask you if you’ve seen a woman with blonde hair, don’t look at her, even though she’s in the lobby and she’s your accomplice. Make sure you look away.” That kind of briefing is very ripe, so that’s been one of the prototypes. The other one has been a lot of the comedy stuff. Things going wrong is ripe.

KS: This is really exciting. Are there antecedents in games for what it is you’re working on? I think there was a pinball game that you controlled with your voice.

RH: There is. There have been a handful of voice controlled games, but very few actual games where you talk and the conversation has been a key part of the experience.

KS: Yeah, that’s very different. Like I wouldn’t say I’m having a conversation with my Xbox when I yell “Xbox on” at it. And even then, it’s a little dicey. I sound like a crazy person most of the time.

RH: It’s funny, isn’t it? It changes the context. How different would it be if the Xbox said back to you “I’m tired because you played Grand Theft Auto too much last night. I need a rest.”

KS: Yeah, if it just said, “No.”

RH: Or if it just said “I don’t like that game,” or “I don’t like the kinds of games you play.” It’s always funny when you get that reaction you don’t expect and that amplifies when you make a game around it.

KS: Do you look at some of your past work differently now that you’re working in voice? Do you look at it like this would have been great for some of the other things that I worked on?

RH: I do. I do. And the AI components as well. It’s kind of interesting. I have to say, just on a personal note, the amount of learning I’m getting here—I think, with the exception of [head of communications] Tom Sarris, I’m the only person here who’s worked in games in the company before—almost everybody else is a Pixar veteran, or people who know how to make characters. So that’s been very educational. The writers here are incredibly talented and when I sit down and listen to them talk about characters, I just learn so much. So, when I’m talking about “Hey, we need this character and I would like this to be the story arc,” and their first question is “Okay, well, what are the flaws? Because the flaws clearly drive the narrative of every character,” and I’m like “Oh, do they? I didn’t know that.” And it’s been kind of great and educational.

///

This interview has been edited for clarity and length.