No matter how much effort you put in to prompting and coaxing it, GPT-3 really couldn't manage anything but the simplest encounter-based game design.
There were things that GPT-3 did really well -- it could write you a 600-word essay on the characteristics and atmosphere of each chamber or passage you find yourself in -- but it couldn't stick to a narrative or engage in anything resembling long-term planning. So, ultimately, you're left with something quite strange and dreamlike: A campaign that's vividly -- even hyper-realistically -- described, but where nothing makes sense and there's no logical continuity over time.
GPT-4 knows what it's doing, though. It's very different. As far as I can tell right now, it combines the outstanding descriptive capabilities of GPT-3 with the ability to (a) remember things it has previously said and establish logical relations over time, and (b) make sure that the campaign appropriately mirrors a "normal" D&D campaign. GPT-3 would frequently mess up and have you fight dragons at Level 1; GPT-4 more appropriately sticks to goblins and bandits. And, when it comes to tactics, pacing, etc., it's really quite good. I suppose it's a little bit conventionally-minded, but that's to be expected.
Almost scary to think that it'll only get better from here on out, because GPT-4 is damn near "good human DM" tier, and already much better than a bad DM.
Edit: but I could be wildly off here?
I've however also spoken to other people, living in different places, and they remember the same time period being all about rules lawyering and "war gaming".
I've also observed a Europe/US divide where the US tends more heavily towards "war gaming" focused role playing with physical maps and moving miniatures around on a table, while Europe tends towards more story focused role playing. But I could be imagining that.
In the 90s and 00s it was almost always about the minis and tactical gameplay. DnD 3.5 was a lot more numbers heavy so it kept a lot of the story people out of the game.
Now DnD 5 the numbers have such a gentle, predictable curve with bounded accuracy and damage and stuff, you may as well ignore them because there's nothing exciting about rolling the dice anymore.
So the story is a lot heavier focus, the character journey. I never used to encounter people who got upset if their character died or groups who had the mentality that characters must always fail forward for the sake of the story.
That stuff is everywhere now.
Just the fact that the tile is "D&D map makers..." instead of "RPG map makers..." is very telling.
But now I don't even see Pathfinder out there much.
I think DnD podcasts and webseries and such were incredibly successful in cementing DnD as the only game in town. They brought in a ton of new players who have only heard of DnD and only want to play DnD.
Here, at least since mid to late 3rd edition, we got a very strong focus on combat "encounters", where you got pre-packaged situations where combat would arise. A combination of monsters, location and possibly events happening during that time. Monsters might even be built for this specific situation, even though their names are generic enough (i.e. a "Orc Battlescout").
Your dungeon FSM gets to a certain state and this encounter gets triggered.
Whereas the canonic ideal of older games was a more fluid setup. Often described as "Combat as War" vs. today's "Combat as Sport". Sure, played very simply, you got an even less sophisticated environment: Room #13 has 4 orcs, guarding a chest, which has 380 copper coins. Kick down door, fight orcs, spend coins on torches for next dungeon. But no experienced DM sees a dungeon that way. Because next to that, you got room #14, where the orc chieftain is preparing the prized pigs they stole from the goblins in room #27, which are currently sneaking up to get it back. Then, as you say, you might have your random encounter, so while the players are heading for the orc room, wolves attack (the goblins aren't guarding the wolf-forest-gates, the DM muses). The fight makes a lot of noise. The orcs look. The orc chieftain looks. The goblins use that distraction. The characters might notice that or not, etc.
Even rather simple setups can achieve rather complicated results, if you're not emulating a 12 year old mastering their first adventure where they just picked monsters from the tables in their Dungeon Masters Guide that are appropriate for a level 2 party of characters.
That's a bare minimum where I'd consider a "AI" DM-like, otherwise you mostly got a NLP interface on gameplay that I could've had with the old SSI Gold Box games on an C64.
I wonder what kind of campaigns we're talking about here. I mean, if it's just mindless modern encounter-based design, sure, I can believe that. Kicking down doors, rolling for initiative, GOTO 10.
But even mildly more moderate DM work like re-jiggering set up monsters to react to the actions of the players (in e.g. a Caves of Chaos type of setup)?