Automatically generated content

(Back to TOC)

5 December 2004

by Mike Rozak

Being a programmer, I feel like I'm wasting my skills when I work on content, like drawing 3D models, designing conversations for specific NPC encounters, etc. I'd much rather be programming, since that's what I'm skilled at... and after all, anyone can create content.

Of course, this last statement is a lie. The truth is, that I know I'm very good at programming, and I know I'm not as good at content creation.

Furthermore, while not anyone can create content, virtually every virtual world will be based on hand-created content. There will be many more hand-crafted virtual worlds than there will be automatically-generated ones because writing all the generation algorithms and dealing with all the unexpected interactions is a very difficult task for a very small market. (I expect that most players will prefer hand-generated content.)

As a programmer, I'd still like to create a world that is 99% computer generated. It just sounds fun... The algorithms to automatically generate content are a fair amount of work to do, but once they're done, the content flows out of them without any effort. (Writing the code to automatically generate an X, such as a tree or dungeon, is usually 10x - 100x as much work as producing a single version of X. So, if a world has 100 or more versions of X, automatically generated X's are a time saver.)

Contemporary virtual worlds already use a bit of automatic content generation, most notably:

• Monsters - An infinite supply of wandering monsters is a no-brainer for automatic content.
  
  
• Resources - Resources like plants, minerals, and even animals are also good candidates for automatic content.
  
  
• Forests - While I could place every one of the 1 billion trees in the world by hand, having a computer do it is much more efficient.
  
  
• Terrain - The terrain in most MMORPGs has at least some element of automation, although it's usually cleaned up by a person.
  
  
• FedEx quests - These are easy enough to automatically generate. Quality quests would be even better, but they're trickier to implement, so MMORPGs haven't yet tackled them.

Of course, I think big, and can go even further:

• Bandits, armies, and police - Automatically-generated monsters are usually solitary. It wouldn't be that much more difficult to make them part of a group roaming the countryside (or city), tyring to meet some randomly-generated goal like "loot all villages" or "capture all looters".
  
  
• Travelling merchants - Same thing as bandits except they don't attack.
  
  
• Villagers - Same as individual monsters, except they don't attack. Various types of villagers could be spawned, from bankers to beggars.
  
  
• Terrain - Realistically, terrain can be created without any human intervention. There are already 3D rendering programs that will create an entire fractal world for you. Rivers, lakes, and oceans follow without thinking. Vegetation can be added based on the amount of water, quality of soil, and elevation.
  
  
• Cities, towns, and farmhouses - These can be generated with a fair amount of coding, and then tied in with the local (automatically generated) villagers. Buildings would probably be chosen from stock models.
  
  
• Buildings - Generating hundreds of variations for houses and other buildings is doable, but fairly difficult to do right. The easiest trick is to create some stock buildings and change their paint jobs and furniture, and maybe a few dimensions. Actually changing their structure involves some trickier algorithms.
  
  
• Underground - Virtual worlds like having caves, crypts, and dungeons for players to explore. These are filled with traps, NPCs, and monsters. While tricky to do, automatic underground environments are entirely possible. (Offline games have been generating random dungeons for decades.)
  
  
• Politics - Cities and nations could come with algorithmically generated politics, which in turn spawn quests. Different factions could (underhandedly) fight for power. Those that win get to put their mark on the city.
  
  
• History - Record the randomly-generated political events and create history.

By the time all these systems have been implemented, I would have a virtual world that is functionally equivalent to most hand-generated MMORPGs, for only a fraction of the cost.

Unforuntately, computer-generated content has a major problem:

• The content, while plentiful, is not as interesting as a handcrafted world. It's like listening to computer-generated music, as opposed to a master composer like Mozart. Computer-generated music is fine for background music, but not for the main experience.

Some people do enjoy playing in automatically generated worlds, such as Rogue, which lets players adventure through a randomly-generated dungeon. Automatically generated content has the following advantages:

• It never runs out - Those people that like playing the same game for hundreds of hours like automatically generated content because of its immortality.
  
  
• Cheaper - Automatic content is cheaper to generate, assuming that enough copies of the automatically-generated objects are used.
  
  
• Players can impact the world - A handmade world cannot allow players to significantly change the world since the changes might invalidate the handmade content, which is very expensive stuff to throw out. Thus, Lord British (in the Ultima series) cannot be killed. If the world were automatically generated, a dead Lord British would be algorithmically replaced (after an algorithmic civil war).
  
  
• A single game/dungeon master - In a table-top role playing game, a single game/dungeon master controls the world for 4-6 players. He not only arbitrates the world's physics, but creates and customises content for the players. A virtual world with thousands of players cannot use this technique since the single GM cannot look after so many players. However, if the GM takes the role of whispering into the ears of the top-level AIs (such as major rulers and enemies), they in turn will dictate to their minions, which eventually produce a better game for the players.
  
  
• The plot object - The plot object, as I described in Story and plot vs. freedom in virtual reality, is much more doable in an algorithmically generated world.
  
  
• Robust - Hand-generated content is "fragile". If you try to do something unexpected to it, it either won't budge (because it doesn't know how to), or it will shatter. Automatically generated content is much more robust.
  
  
• Uniqueness - Knowing that the world as you're experiencing it only exists on your computer, and is not replicated on millions of users PCs, is somehow gratifying.
  
  
• Feels more real - A hand-generated world doesn't feel real (to some people) because of all the above problems. For most people the unreality of hand-generated content is a non-issue, and perhaps even desirable. Some people find it very distracting though; I remember when I played with lead miniatures as a teenager, I never liked characters that had built-in stands. The stands were there for support, but the fact that a piece of ground moved around with the character was disturbing.

Boiling down the list, it's clear that automatically generated worlds attract the following people:

• Lots of time and not enough money to pay for hand generated content. If this is the case, would these people chose to play an automatically-generated virtual world that lasts 1000 hours, or 10 free 100-hour hand-crafted worlds?
  
  
• People that want to have an impact on the world, even if it is only a virtual one. These people may also be attracted to PvP worlds.
  
  
• Those looking for a more realistic virtual-world experience. Would these same people also be attracted to 3D graphical worlds instead of text worlds since 3D graphics are more realistic?
  
  
• Players wishing a personalised experience will be attracted to automatically generated worlds. Knowing that the members of a village solidly support your player, or that a king has sent out assassins to kill your player provide a more meaningful experience.

I suppose that, in the end, it's all a tradeoff, and the decision depends upon what you're interested in, and what kinds of users you wish to attract.