Richard Bartle Q&A log
The full log of a great Q&A session with Richard Bartle in Metaplace has been posted up on the Metaplace Forums. It was a wide-ranging discussion, attended by over 70 people. Richard’s dry wit was, as usual, on full display.
A typical, provocative, snippet:
[05/26/09 13:13:10] gguillotte: I’ve been watching procedurally generated content for a while. Love comes to mind, a PG MMO. What sort of impact is this going to have, where content generation is automated?
[05/26/09 13:13:45] Richard: it depends if the generation of the content is the game or is filler
[05/26/09 13:14:11] Richard: procedural content can work – I’ve spent many, many hours playing Rogue for example
[05/26/09 13:14:42] Richard: using procedural content to create a canvas for virtual worlds seems a perfectly rational thing to do
[05/26/09 13:15:22] Richard: however, the designer has to put their soul in it somewhere: either this is by modifying the procedural content or by creating the framework that creates it
[05/26/09 13:15:59] Richard: now the former is the traditional way for designers to speak to players; if a designer wants to speak through the content-generation rules, well
[05/26/09 13:16:12] Richard: that would be possible but we don’t have the vocabulary for it yet[05/26/09 13:16:28] gguillotte: Thanks.
[05/26/09 13:16:31] Richard: that makes it an interesting time for us
[05/26/09 13:16:38] gguillotte: Indeed 😀
[05/26/09 13:17:11] Richard: Metaplace is a similar thing, btw – we’ll see things here that we haven’t seen the like of before
[05/26/09 13:17:21] Cuppycake: (We already have!)
[05/26/09 13:17:24] Richard: which is why I’m so enthusiastic for it
[05/26/09 13:17:55] Richard: I don’t mean new worlds, I mean new ways of communicating through world creation
21 Responses to “Richard Bartle Q&A log”
Sorry, the comment form is closed at this time.
Voice fonts, I would think, would bring out the roleplayer in players. Set the tone, so to speak, of the game.
I’d also like to see voice commands, including for emotes.
Realism can make a game much more interesting. It goes without saying that there has to be loads of fun, too. But adding realism such as getting wet and ruining things that would be ruined by water, this just makes the game world feel like it’s more real. Immersion. I’d love to have to prepare properly for long treks in wilderness, in search of lost temples and ruins. This would make such efforts more challenging, yet doesn’t have to affect other, more typical, game play.
I didn’t much fancy the voice font idea. It’s hard enough for people who don’t speak English as a native language to keep up with voice chat as it is without deliberately inserting weird accents.
Also, I’m puzzled by the weight he puts on pseudonymity. Most people who use voice chat do it because they like to listen to each other’s real voices (because they’re friends, or becoming friends).
I didn’t manage to go into depth on many topics, which is probably just as well given that there were people in the audience who know more about those topics than I do. I also didn’t manage to answer all the questions.
Part of the reason for the latter was that I couldn’t expand the chat box to fill the majority of my screen, so questions disappeared before I got back to them. Also, because there was no way for people to tell whether I was typing or not, they didn’t know if I’d finished my answer. Back in the days when we had this kind of thing in CompuServe, the way it worked was that whoever was asking/replying would end their message in ga (for “go ahead”), but this only works for formal Q&A sessions, not ones where people can chip in with remarks from the floor.
The other thing CompuServe did was to use a special channel. One person was the moderator, other people (the speakers) were given full permission to speak on the channel, and everyone else who used the channel had their questions queued up. The moderator would then release them one at a time when each had been answered, ignoring those that were already answered or off topic or badly put or whatever. With an audience of 300 or so, this would probably be necessary; it wasn’t needed this time round though (we didn’t even need bouncers).
The other thing, which I’m sure the Metaplace folk noticed, was that with so many people in close proximity it’s hard to see who is whom. Their name badges all overlap. It was handy when I did see them to note what level the people speaking were, though, so I knew how scared of them to be!
All in all, it was quite fun! I dare say you (Raph) could have answered every question I was asked better than I did, but you’d have needed a bigger auditorium to house everyone who wanted to come. As an experiment as to what format works, though, on the whole I’d say it was a success.
Richard
I’m thinking of the old writer’s trick of putting story elements on index cards, then shuffling and pulling out characters, settings and plot points at random. In this case, each discrete element of the story is created (or at least chosen) by the author even though the particular combination of elements is random.
I think the vocabulary is going to resemble those writer’s techniques, improvisational comedy, or a jam session. We need algorithms that can take handcrafted elements and fit them together on the fly into a balanced, aesthetic whole with seamless transitions and logical connections.
Combine with code that monitors and analyzes what the players find interesting, what they focus on, what they always come back to, and which has the capability to expand upon those aspects.
For extra credit, include channels by which players can submit their own elements for evaluation and possible inclusion.
Perhaps the greatest statement that a designer could make would be to design something so evolutionary, so able to grow and adapt, that it becomes much more than any one person could ever envision.
We actually do have a moderation system with message queue, multiple channel support (so you could have a backchannel), and a no-nametag plug-in, set up in the The Stage. But there was some confusion about where the event was to be held, and we ended up outside where none of those facilities exist.
I think this is a dangerous assumption. It may be true in social worlds, but in game worlds, the commonest reason to use voice chat is because of the operational advantage it gives: it’s faster to coordinate complex game maneuvers with voice.
@yukon sam: We used to call that “slice and dice” and it is still a technique for generative creativity with a healthy dose of serendipity. In another way, it is just loops and while they are a permanent tool in the songwriter’s toolkit, most have backed away from them outside certain genre where beat is what matters. My son has been ‘generating’ beds from his synths so he can improvise on trumpet over them. In fact, the kids are working on some wild, weird but germinal styles mashing together electronica and trad jazz chops. Tools is tools.
@raph: I dislike voice chat unless proximity filtering is included. Compressors chop the consonants and that increases the time required to interpret whereas when reading, we can chunk larger phrases faster. Given a team conditioned for imperatives, you’re right, Raph, but I’m not convinced it is true in all non-operational situations. On the other hand, as I pointed out elsewhere, audio streaming gives Second Life a big advantage because in-world musicians can play live. Otherwise, I’m starting to wonder if Big Worlds (say lots of terrain) really are advantageous or that useful unless questing and roaming are. IOW, the Vivaty guys may have a real advantage simply because they are embedded in the social network and social events favor small rooms.
Small may be beautiful.
Yukon Sam said on May 27th, 2009 at 7:02 am:
I think the whole idea of procedurally-generating stories is fundamentally misguided. By all means develop a vocabulary for the story-physics of a world (eg there are players, that can accept quests from quest vendors, and can turn in quest tokens, etc). Procedurally fill the world with important and urgent things for the players to do and run enough of a world sim to keep it filled. But story belongs at the end. Story is what you have when you look back on your actions and create (or have created for you) a narrative that ties all the actions together. For that you need a pattern-matching story engine, not a procedural generator.
For example, a story-engine matches the facts that a character has a bad relationship with a sibling but a good one with a parent and adds a “Mother always liked you best” story to the character’s narrative. The procedural/sim world happens to result in the sibling getting in trouble and the character quests for the mother and saves the sibling. The story-engine matches those facts and adds a “Dutiful son” story to the character’s narrative. The character fails to repair the sibling relationship and the world sim happens to simulate the sibling betraying the character. The story-engine matches the “mother always liked you best”, “dutiful son” and betrayal facts and adds a “No good deed goes unpunished” story to the character’s narrative.
Players generally have more fun when they are able to live in the moment, whether that be fighting, questing, socializing, crafting, etc. Players love to tell stories after the fact about “that time I mispulled a demon and we all wiped”, “the time we almost wiped but I saved us with a clever use of skills”, “the masterwork sword I created and donated to our guild’s best warrior”, “that one night we organized a conga line through the city”, etc. The best stories put the focus on the player, are unique to that player, provide narrative closure to a series of past actions, and are worth retelling. Ordinary quests fail on most counts. A pattern-matching story-engine has the potential to do so much more.
I remember hearing of a WoW raiding guild that got voice for the first time. They fired up chat and got ready to head out… only to find that their trusted veteran leader of many campaigns sounded very much like a 12-year-old boy (because, as it happens, that’s what he was).
What changed? The character was still a mighty warrior, and the player had not changed a whit — he was still a canny campaigner with the same leadership abilities. But once his voice was heard, the respect he once commanded evaporated, because the preconceptions and prejudices of the other players were stronger than their own personal experiences and observations of the leader’s skill.
Similar experiences with voice have been reported by players with strong regional accents and by women in command positions. They are often marginalized, disregarded or relegated to secondary positions just because they don’t sound “right” as leaders… “right” in most cases being a baritone American male.
If you must have voice in a game (and it does help a lot in coordinating group combat) it would indeed be helpful to have the same sort of control over your aural appearance as you have over your visual appearance. It’s all part of the same package.
There’s this skill called voice acting… It’s part of roleplaying. And one could argue that voice acting is a metagame. When elements of performance are incongruous, the spell of immersion is broken. Many players, unfortunately, aren’t conscious that they are roleplaying because the system is in charge of animation and wardrobe.
@yukon and morgan: That may be why voice synths work as well as they do. They can be fit to the character.
Rats! The Voice Props: yet another set to configure and share off the server. 🙂
But basically, the more real and personal we make it, the faster the web finds out we are dogs. That argues against integrating social webs into roleplaying games. It is The Guild. Reality is intrusive.
len, we do audio streaming too. 🙂 And yeah, small can be beautiful for sure. I am not convinced voice is always better in non-operational situations either. Often it is worse, for many of the reasons others have cited.
@raph: But of course you do. I would expect nothing less.
The “small is beautiful” meme keeps nibbling at the edge of my thoughts. Some, particularly Forterra, made claims that very large terrain handling was the sine qua non for 3D web standards. I am beginning to question if it is so. It seems to me that rich locales are more important than flying over lots of terrain with some profitable but not very ubiquitous exceptions. Maybe the *Metaverse* is a Shaggy Dog story and ultra-realism is for photographs and video.
BTW: off topic but because we both noodle on our songs:
http://home.hiwaay.net/~cbullard/mp3/BeautifulBoo.mp3
At 1:45 approx, one of my shorter ones but my holiday distraction. Who says songs should be longer when all they really are is more?
Sherwood makes heavy use of procedural island and dungeon generation, much in the same fashion as Rogue. As a tiny company (just my wife and I) procedural generator are vital to getting out of the content trap. I don’t have an army of 3D artists to do another pass every time I make a change. Richard’s choice of words using “procedural content to create a canvas” is right on. By using a master seed for an island or dungeon level, you can start by testing different seeds and parameters to see if some interesting comes out auto-magically. You’re seaching for lucky accidents. Then you can add traditional world design or narrative on top of that canvas. This saves massive amounts of time relative to starting from a blank page. Updates you make to the generators propogate instantly through the level without requiring another labor intensive art pass. It takes a while to get used to, but as the generators gets more sophisticated, new content becomed easier and easier to create quickly. The other benefit for a browser based 3D MMO is that content generated at runtime is VERY small relative to traditional content.
There are several different ways of changing someone’s voice:
– The simplest is ye-olde pitch and formant shift. It doesn’t change your accent. It can make you sound like a midget or a giant. As a general rule, the more the voice changes, the more unnatural it sounds.
– Voice transformation using HMMs is the next in line. It transforms the voice without understanding what words (or phonemes) are being spoken. It’s more convincing than pitch and formant shift, but all the samples that I heard had quite a lot of distortion. (Google for hmm voice transformation)
– Do speech recognition followed by text-to-speech, keeping the prosody (pitch, duration, volume) from the original speech. This should work, except that speech recognition isn’t terribly accurate, especially when you’re screaming at it (“Get out of the way! It’s a dragon!” might become “Sit out the day! Pizza wagon!”… spoken in an authentic scottish accent.)
– Text to speech with typed in text. Works, but (a) it doesn’t convey emotion well, and (b) typing text is a nuissance. I believe that one TTS engine hooks itself into Vivox and does this.
Shameless plug for someone else: If you want to hear the latest text-to-speech technology from a number of research organizations, please take the http://groups.inf.ed.ac.uk/blizzard/blizzard2009/english/register-ER.html listening test. In each section, you’ll hear a sentence from each of the 17(?) participating research organization and be asked to rate them. One of the recordings is real speech. A few of the text-to-speech engines are very good.
I think it is dangerous to “conflate” two different use settings. What is and isn’t appropriate and desirable when it comes to voice chat is probably cultural, not a given. I.e. Voice wasn’t common in Anarchy-Online when I studied it, but it was sometimes used for social reasons in teams of friends, some groups of friends used it by default, some only occasionally. Obviously some people take on new identities in MMOS and would not welcome voice for that reason alone, but I believe those are a small minority… My guess is that the inconvinience of the technology would be more important for most people (impractical / shy or insecure players).
Generating an entire landscape from a single seed is pretty cool. What would be killer (and unfortunately, mathematically improbable) would be the ability to take a scene rendered by an artist and reduce it losslessly to a seed.
While I believe the computing power needed to crunch numbers in this manner would literally be astronomical, it might be practical to match a building, landscape or maze to a preset form that can be easily generated by a seed. Then instead of transmitting the entire object to be rendered, you would just transmit the seed and an exception list to modify the preset.
Free up some pipe and a lot of really interesting things become possible. It’s not that I don’t think small is beautiful; I just think the dominant paradigm for this century is going to be “there is no dominant paradigm”, which leaves plenty of room for all creations great and small.
And I’m sure that anybody actually familiar with the topic is going, “well duh, way to reinvent the wheel, Ace” or “been there, done that, didn’t work”. Which would explain why I’m still coding boring business apps in VB for clerk wages:{
All of the landscapes in Star Wars Galaxies were made with a tool that did much like what you describe, Yukon Sam.
In addition, there’s a company that sells tools for exactly this purpose for texturing 3d models: Allegorithmic.
Oh good, I’m only nine years behind the curve on this one 😀
@Yukon: yeah, it is mathematically impossible, because terrain is noise and noise cannot be compressed… However, if you accept lossy compression then there has been some research on fractal compression… Still, I think you’ll be better off letting the artists work with procedural modelling tools as Gene Endrody seems to do with his worlds.
You can do lot of interesting things with collages of iterated function systems/L-systems. If you want variation you need to set up more than a single generator though (mandelbrot tends to look like mandelbrot) and combine them in various ways.
Exception lists for dynamic worlds only work if you accept that the world is metamorph (e.g. allow trees to move about), otherwise the exception list will eventually grow larger than the world itself… So a “seeded approach” has serious design/architectural implications.
Anyway, graphical artistic programming languages might provide intersting starting points: Context Free Art and Processing. IMO, if you allow uploading of compressed landscape-generating code to the client then you can have more variety and interesting landscapes than by just sending seeds/DNA. And it is more difficult to hack… (Client is in the hand of the enemy etc…) Just look at what people have done with solid modelling in povray, and povray source files tend to be small.
Ola, thanks for the resources. I’ve had an interest in the topic since the day I got bored at work and rigged Excel to generate random landforms with graphing tools. Looks like I’ve got some catching up to do 🙂