Kami Harbinger

Raph Koster just posted a reprint of Whither Online (circa 2005), where he said:

Anyone remember what cyberspace looked like a decade ago? There we were, all fresh arrivals in the Metaverse, dreaming of Snow Crash’s virtual bars and William Gibson’s skies like televisions on dead channels. We wondered if the Holodeck would require one of those newfangled 3d hardware video cards or not. If we were really old-school sci-fi fans, maybe we thought about Bradbury’s Veldt or Vernor Vinge’s "True Names."

Back then, we dreamed about dynamic worlds that we could morph on the fly with a thought (or at least a twitch of a mouse). We had a lot of grand visions about really realistic NPCs that would move about the world like the ones in Ultima VII did. We thought maybe the orcs would be invading virtual towns because they wanted to, not because there were spawn points set up by the city gates.

Then he goes on to talk a lot about the failed dream of AI and artificial life generating quests that people would like. This is the wrong answer. Putting AI in an MMO is like sitting in a restaurant surrounded by free gourmet food, but wishing you had a McDonald's shitburger.

As always, I point at Second Life in answer. Even in 2005, this should have been the obvious answer, but 2 years later, it's not really in question.

We already have a metaverse where there are hundreds of thousands of ongoing melodramatic plots created by mostly-intelligent agents. The good kind of agents, real people, not stupid AI NPCs. AI will never work and people don't like NPCs. Programmers tend to like NPCs because they're easy to set up and control, but actual players, even those programmers when they're in the game, despise them, because they're just machines. There is only one NPC I've ever felt any empathy for in 30+ years of playing computer games: Floyd from Planetfall. And I was maybe 12 when I played Planetfall. The use of NPCs is a delusional sickness brought on by spending too much time alone with your computer.

I'm not talking here about "monsters", which are merely a mechanical obstacle like a puzzle or a locked door. It's revealing that when MMORPG players talk about killing monsters over and over to get experience and material rewards, they call it "farming", not "hunting". Monsters are psychologically equivalent to plants, not animals or people. An NPC is an artificial person you're expected to interact with and give enough of a shit about (if only for the reward or reputation boost) to carry out a quest. But in MMORPGs, there are no bad consequences for failing to carry out an NPC quest. If I ignore a "sick" NPC and don't bring her medicine, she'll still be there tomorrow. If I do heal her, she'll still be there tomorrow, coughing away for someone else. What do I care? There can be no world-altering alternative, because it doesn't scale past one player.

True AI was a boondoggle. When I was younger, I believed it was possible, but as years passed and no progress was made, and I studied the issue more, it became obvious that it will never work, and that any attempt to make it work is doomed to failure. Worse, it's stupidly wasteful, because we already have 6 billion mostly-intelligent humans on this planet. We don't need non-human intelligences to put characters in games.

Artificial life looked like a promising field for gaming, initially. But it's too unpredictable, and too fragile. You can set up a field of creatures or plants, and hope that they'll expand fast enough to counter demand, but very quickly the players will plow into them in such numbers that the tipping point is reached, and they go extinct. If you do set them up to reproduce fast enough to counter the number of players, then the slightest lag in harvesting will have them swarm over and destroy everything. It's massively unstable. I'm appalled that Raph still believes in this nonsense; he experienced first-hand the total failure of his first artificial life design for Ultima Online, and apparently learned nothing from it. It wasn't just an economic and ecological catastrophe, it rapidly became boring except for the Player vs. Player element. Real human players provided the only challenge in that world.

You cannot have designer-prepared plots or AI-prepared plots for more than one player in a mutable world. It's a logical impossibility. Suppose you get things set up just right so that just one player can try to carry out a specific planned quest; someone else will be disrupting parts of it elsewhere, and the player will fail. The more quests you try to set up and juggle at once, the more chaotic the interactions get, and the more disastrous the failure.

There is a way out, though, by accepting that people exist, not fighting it. The plot in Second Life is pretty compelling if you like people at all. It's people going out and having fun (and sometimes having conflicting ideas of what "fun" is), and running businesses and succeeding or failing, and falling in love, and falling out of love. Because it's just real life with fewer consequences, and accelerated to Benny Hill speeds, it has all of the bizarre plot twists of a daytime soap opera, but you get to be the star of your own show. Everyone does.

Any RPG quest type has an equivalent, you just have to be more creative. Consider the standard RPG "supply and demand" quest, where someone needs a resource and someone else supplies it. I make good pocket money as a skilled craftsman, the SL equivalent of a small electronics shopkeeper, making gadgets that people ask for. It's not just farming crystals and materials and then clicking a button, either; I have to build prims and write scripts and test my software. It takes actual skill and interest on my part, it's not a repetitive activity. That's been a fun and both personally and financially rewarding quest. If you can't script, you can make clothes or buildings. If you don't want to build (though I think that almost everyone will want to build things eventually), you can run a tour group, or host a fan-club, or strip, or whatever.

The artistic urge that everyone has, the drive to have something to do besides just talk, has created, from a blank canvas, a nigh-unlimited range of games and quests and social activities. If you like swordfighting, go to Samurai Island. If you like casual and gambling games, go to a casino/arcade. If you like porn, there's strip clubs. If you want to talk Star Trek fandom, there are Star Trek fan-clubs. If you like live music concerts, there are always multiple performers in SL at any time. If you want to pretend to be a furry, or a vampire, or whatever makes you happy, there are 24/7 roleplay sims for that. If you like exploring, Second Life is crammed full of interesting things people have built. Just pick a direction and start flying, and BlogHUD anything neat you find.

Everyone on the planet can't participate in SL quite yet. It needs more servers, and it needs more efficient use of those servers. While the peak concurrency has more than doubled in the last year, it has years to go before even a million people can be in SL at once (circa 2011, at this rate). LL needs to learn how to load-balance and transfer regions across servers; server-bound regions are not a scalable solution. Probably it'll have to wait a couple years until the server can be open-sourced before SL truly becomes the Stephensonian Metaverse. Linden Lab is going to have to transition from sole proprietor to protocol definer, region and user registrar, and banking authority. There's more to do on the user side, as well: more activities, more interesting sims. Wasting time on ideas we know will fail is folly. Working on activities and tools that enable real people to communicate and have fun together, that's what we need now.

The Metaverse is here, but that's just the start.