Some of my favourite games use procedurally generated maps. But, those maps are not hand-sculpted the way MMO dungeons are. And, while you could certainly use generative AI to come up with generic babble from NPCs, that’s not the same as designing entire quests. It may be that eventually a generative AI system will be able to do everything a human could have done: hand-crafted maps, full quest chain dialogue, etc. I just think we’re nowhere near that point yet.
For example, a quest chain almost always has a goal behind it. You’re revealing a certain aspect of the story to the player bit by bit as they complete parts of the quest. But, to do that you need at least a very basic theory of mind. You need to understand what the player knows before the quest chain starts, what each bit of the quest chain will add to their knowledge, and then what they’ll understand at the end of the quest chain. That “theory of mind” stuff is the thing that generative systems just can’t do right now because they’re just fancy auto-complete.
As for auto-generated dungeons, WoW tried that with Torghast in the Shadowlands expansion, and it was not well received. Granted, part of the problem was that Torghast was a depressing, death-themed “dungeon”. But, a bigger issue was that there was no intention behind the design of the levels. It was just a randomized set of corridors that fit together in a random way. Good dungeon designs require intention. You want to reveal something to the player as they go through the dungeon. Ideally you want to know that you’re working your way towards a boss. WoW’s black temple raid is a good example of this. You start in the sewers, you work your way out into a courtyard, you enter another building, clear out the ground floor and open a door that unlocks access to a set of staircases that works its way to the top of the building. You beat the Illidari council which allows you to access a door that opens to the roof of the building where you face the final boss Illidan. I don’t think generative AI is anywhere near being able to come up with a concept like that, let alone design the maps and art for the whole thing.
I wasn’t thinking of an AI generating the world map and dungeon. I was more thinking of an AI driving the agency of world actors. It doesn’t have to have a complete theory of mind, a reactive AI or limited memory AI, aka “chess engine” could “simply” drive the opposing faction.
We could imagine a war scenario, where the AI plays one side and the players the opposite and the effect of war would naturally change the world. That town where the wiki tells you you could buy that cute horse? Well too bad the AI invaded it or reduced it to rubbles. The HQ where the commander is supposed to be, well it moved back 20 clicks after the last player attack, etc…
The AI doesn’t really need to understand the purpose of its objective.
Of course it’s a frigging huge undertaking, it would probably cost the GDP of a small country and need it’s own nuclear powerplant to run (hello Microsoft!) but not impossible with today’s tech.
That’s one thing I’ve always admired about Eve Online. It’s an MMO that’s almost entirely player driven. Various sectors of space change hands between different factions of players. That results in the sorts of things you’re talking about. Unfortunately Eve has extremely boring space battles (for players, for watchers it can be fun), and a toxic community.
But, I’ve always wanted an RPG where the world evolved. To me, the key thing to make that realistic would be NPCs that didn’t respawn. Like, if you killed a certain golden dragon named Gurnadom, that dragon was dead, gone, nobody else could kill it. There would be no Gurnadom killing guides because there was only ever one Gurnadom and only one group of players ever killed that dragon. There might be tips on killing golden dragons, but each dragon was unique so it wasn’t a matter of watching videos and understanding the patterns. Each fight against a golden dragon could only happen at most once, and every fight was unique.
And, in any game involving war, there should be permanent destruction of things: fortresses that were attacked would take damage over time and eventually be turned into rubble. A side that’s winning a war should be expanding its territory. As a result, where a player can safely go should depend on the progress of the war, which is something not programmed into the game, but player driven.
I’m just so tired of the WoW style of MMO where the player is “The Champion” who has saved the world multiple times… along with the hundreds of other nearby players who are all the one-and-only champion who also killed a certain raid boss over and over every week for a month.
Yes, I totally and entirely agree with all of that. I also would love to see permanent impact of actions on the world and ditch player is the chosen one, hero of the ages paradigm like the 2k other players.
Let design a game together and we’ll just need a 100 millions $ to produce it :D
Some of my favourite games use procedurally generated maps. But, those maps are not hand-sculpted the way MMO dungeons are. And, while you could certainly use generative AI to come up with generic babble from NPCs, that’s not the same as designing entire quests. It may be that eventually a generative AI system will be able to do everything a human could have done: hand-crafted maps, full quest chain dialogue, etc. I just think we’re nowhere near that point yet.
For example, a quest chain almost always has a goal behind it. You’re revealing a certain aspect of the story to the player bit by bit as they complete parts of the quest. But, to do that you need at least a very basic theory of mind. You need to understand what the player knows before the quest chain starts, what each bit of the quest chain will add to their knowledge, and then what they’ll understand at the end of the quest chain. That “theory of mind” stuff is the thing that generative systems just can’t do right now because they’re just fancy auto-complete.
As for auto-generated dungeons, WoW tried that with Torghast in the Shadowlands expansion, and it was not well received. Granted, part of the problem was that Torghast was a depressing, death-themed “dungeon”. But, a bigger issue was that there was no intention behind the design of the levels. It was just a randomized set of corridors that fit together in a random way. Good dungeon designs require intention. You want to reveal something to the player as they go through the dungeon. Ideally you want to know that you’re working your way towards a boss. WoW’s black temple raid is a good example of this. You start in the sewers, you work your way out into a courtyard, you enter another building, clear out the ground floor and open a door that unlocks access to a set of staircases that works its way to the top of the building. You beat the Illidari council which allows you to access a door that opens to the roof of the building where you face the final boss Illidan. I don’t think generative AI is anywhere near being able to come up with a concept like that, let alone design the maps and art for the whole thing.
I wasn’t thinking of an AI generating the world map and dungeon. I was more thinking of an AI driving the agency of world actors. It doesn’t have to have a complete theory of mind, a reactive AI or limited memory AI, aka “chess engine” could “simply” drive the opposing faction.
We could imagine a war scenario, where the AI plays one side and the players the opposite and the effect of war would naturally change the world. That town where the wiki tells you you could buy that cute horse? Well too bad the AI invaded it or reduced it to rubbles. The HQ where the commander is supposed to be, well it moved back 20 clicks after the last player attack, etc…
The AI doesn’t really need to understand the purpose of its objective.
Of course it’s a frigging huge undertaking, it would probably cost the GDP of a small country and need it’s own nuclear powerplant to run (hello Microsoft!) but not impossible with today’s tech.
That’s one thing I’ve always admired about Eve Online. It’s an MMO that’s almost entirely player driven. Various sectors of space change hands between different factions of players. That results in the sorts of things you’re talking about. Unfortunately Eve has extremely boring space battles (for players, for watchers it can be fun), and a toxic community.
But, I’ve always wanted an RPG where the world evolved. To me, the key thing to make that realistic would be NPCs that didn’t respawn. Like, if you killed a certain golden dragon named Gurnadom, that dragon was dead, gone, nobody else could kill it. There would be no Gurnadom killing guides because there was only ever one Gurnadom and only one group of players ever killed that dragon. There might be tips on killing golden dragons, but each dragon was unique so it wasn’t a matter of watching videos and understanding the patterns. Each fight against a golden dragon could only happen at most once, and every fight was unique.
And, in any game involving war, there should be permanent destruction of things: fortresses that were attacked would take damage over time and eventually be turned into rubble. A side that’s winning a war should be expanding its territory. As a result, where a player can safely go should depend on the progress of the war, which is something not programmed into the game, but player driven.
I’m just so tired of the WoW style of MMO where the player is “The Champion” who has saved the world multiple times… along with the hundreds of other nearby players who are all the one-and-only champion who also killed a certain raid boss over and over every week for a month.
Oh and I never actually tried Eve. It was great to read summaries of the big events but actually playing it seemed more like a 9-5 job than a game.
I remember reading, a “how-to start in eve” and thinking “hell no, I’m already doing that shit in the office”.
Yes, I totally and entirely agree with all of that. I also would love to see permanent impact of actions on the world and ditch player is the chosen one, hero of the ages paradigm like the 2k other players.
Let design a game together and we’ll just need a 100 millions $ to produce it :D