Saturday!
My buddy Professor Esperanto wrote me with a link about Artificial Intelligence and D&D.
While the post itself asks the rather interesting question: What would it mean for an AI to ‘pass’ as human in a game of D&D? Instead of the Turing test, perhaps we need an elf ranger test?
The article itself goes on to discuss such things as embodied intelligence and social intelligence and the importance of making sure that future AIs aren’t merely optimizing on one axis of what we understand intelligence to be.
But my immediate thoughts were, instead, to articles like this one that talk about the new AIs learning to play Go and how they are, and I’m quoting this here:
“They’re how I imagine games from far in the future,” Shi Yue, a top Go player from China, has told the press. A Go enthusiast named Jonathan Hop who’s been reviewing the games on YouTube calls the AlphaGo-versus-AlphaGo face-offs “Go from an alternate dimension.” From all accounts, one gets the sense that an alien civilization has dropped a cryptic guidebook in our midst: a manual that’s brilliant—or at least, the parts of it we can understand.
Which, instead of making me think of what we could do to learn from an AI opponent as we played more and more games against it, made me instead think of an AI *DUNGEON MASTER*.
Now, we’ve all played a D&D game with a DM who was a bit of a jerkface who wanted a TPK before the end of the night. But imagine a DM who, instead, gave us a dungeon that it had played against itself 10,000 times before bringing it to the table.
A DM who put together Neverwinter Nights III and wanted us to play it first as a Monk, then as a Cleric, then as a Thief, then as a Barbarian.
Games like Chess and Go are just the beginning. Games where an AI can play in the campaign are the next step… but the step after that is the games where the AI is the DM… and the DM understands that the goal is not, necessarily, a TPK.
Unlike that guy we used to play with.
So… what are you playing?
(Picture is HG Wells playing a war game from Illustrated London News (25 January 1913[/efn_note]
I mentioned Left 4 Dead recently. I really like its AI. The theme of the game is that it’s a zombie movie – it follows all the tropes, has music that swells during the big fight scenes, and even ends with movie-type credits. The “movie” has four main characters, and if you play solo, the AI takes three of them, and does a decent job with them. The other characters’ kill stats are comparable to an average online player’s. There’s also an AI system called The Director, which paces out the game very nicely. The game doesn’t have preset boss fights. The Director keeps good pace with the players, and sends in swarms and/or bosses just often enough to keep things competitive. It doesn’t exactly let up on you when you’re in trouble, but it will never let you get too comfortable. It really does feel like a good DM.Report
Long ago, I was into this random dungeon generation thing.
It started as a way to move things in a continuous story line, rather than “… and after you come out of that dungeon, everyone levels up, is at peak hits points, and then, all of a sudden, the party is right in front of this other dungeon . . . ”
I came to understand “base of operations,” travel to and from the dungeon, and going from one town to a neighboring town. This stuff is good for lightening up the characters who want to take 10,000 gp with them.
Anyway, in the travel from one town to another scenario, they can come across all kinds of things. First came the orc patrols, then the bandit clans, etc. At last, they come across mini-dungeons of an impromptu type.
The tables I had originated from the appendices of the 1st ed. DM Guide, but it split off in a big way, in that there were different tables for different types of dungeons; e.g., a maze of caves, ruins of an old castle, etc.
I always thought that would lend itself well to AI generation. The tables require human interaction to resolve conflicts judiciously. AI could learn the best way of doing that, maybe even better than me.Report