In reading over the ridiculously overhyped gamer’s manifesto from last week, Jamie points out what I failed to, which is that people think they want realistic AI, but they really don’t. His quote:

One of the reasons we make game AI stupid is not always lack of processor power but often because a smarter AI would school our players too deeply and they’d turn to simpler games. Look at chess – we’ve got some great AI there, so great nobody wants to play against the computer. Stealth games would be another example: if the security in a stealth game actually was decent, playing these games would become an exercise in frustration instead of an exercise in feeling like a kick-ass ninja commando.

Yep. It is ridiculously easy to write an AI that will kill you instantly.* And you may think you want a world where every guard in the complex converges on you if you trip an alarm, but in reality, you don’t. It is actually a much larger challenge to write an AI that doesn’t pwn on demand, but still appears intelligent.

A lot has been made about classic EQ-style AI, and how combat classes develop around it (crowd control, puller, etc). But the fact of the matter is that realistic AI on an MMO — mobs always attacking you when they see you, mobs always bringing friends, mobs not being easily distracted by the shiny object while you open a can of whoopass on their friend — would result in players besieged in their little towns, unable to leave the city walls in groups smaller than 50. This is provided the server doing all these AI calculations doesn’t let out a muffled cry before exploding into flame.

In most games, the best way to think about A.I. is to think about it as a puzzle the player needs to solve. It’s roughly analogous to the cards that start face up in solitaire – it’s the visible circumstance that you have to modify your gameplay patterns to deal with. Everquest’s model has persisted because it provides an interesting problem that can be solved either by a solo player or a group.

EQ’s AI is also somewhat deterministic, which is to say that if you know how the AI works, you’re not going to be surprised a whole lot. You certainly could have more randomness thrown in and/or hide more behind the virtual DM’s screen, but having a game where the AI is fairly transparent makes it easier for groups to congeal and plan their strategy. The more the AI requires players to evolve to random circumstance, the less faith players will have in casual pickup groups.

This doesn’t mean that EQ is the only way to build MMO AI – quite the opposite, I hope people experiment with alternatives to it quite a bit. It also doesn’t mean we shouldn’t strive for elements of AI that aren’t realistic – in Half-Life, it was quite amazing when the enemy units started shooting from cover. But (and this is important) it was beatable. And ultimately, AI that provides an interesting yet beatable problem is the goal.

———————
*Turn-based strategy games are the exception. Top notch computer engineers have spent 30 years trying to write a chess AI that can reliably beat the masters with a supercomputer’s worth of CPU cycles, and yet you expect an AI that can beat you at Civ without cheating on your gerbil-powered laptop?

Original comment thread is here.