The Real Challenge Is Teaching AI to Play D&D
wired.com
wired.com
[0] https://en.wikipedia.org/wiki/Paranoia_(role-playing_game)
There an AP of paranoia with Felica Day as a play that woman is scary good.
Plus, we already have (weak) AI DMs. They're called video game RPGs. You're the player (as are many others in the case of an MMO), the game itself is the GM. They do many things better than a human could (handle lots of background math, present gorgeously rendered environments, have professionally voice acted NPCs, have skill based real time combat), but the audiences don't entirely overlap because TTRPG players are seeking the freedom of choice and limitless outcome that video games cannot really match.
Personally, I say let things fall to the medium that best suits them. If D&D is sufficiently complicated rules wise that it is actually more fun to have software take care of that so you can just focus on the tactical combat and scripted story choices, then the video game is probably your best bet! There are many other TTRPGs where this is not the case, and where any digitization would take away much more than it would add. Such RPGs are elegant enough rules wise that they do not slow down play at the table, and you can spend 99% of your time just relishing each other's madcap plans and imaginative descriptions of things.
[0] https://www.polygon.com/2020/2/27/21156082/baldurs-gate-3-di...
These kind of stuff never happens in CRPGs - it couldn't because there's infinite number of possible open-ended solutions to any problem. If AI can improvise them I'd argue it's as good as a Turing Test.
BTW complicated rules aren't that big of a problem - there is a compromise between CRPGs and TTRPGs - and it's computer-assisted TTRPGs played over internet. The most important advantage is that you can play with people from other continent, so it's much easier to find players. But the computer-assisted part is also making the rules much less of a problem. You set up your character (all the items, feats, skills, abilities) before the game, and when DM wants he asks you to roll the attack, you click and the system rolls the dices for everybody to see, adds the needed modifiers, calculates the damage (including critical hit if needed etc. ) and shows how much hitpoints you lost.
It's faster than doing all this stuff physically and beyond the initial setup and occasional level-up it makes complicated rules more bearable.
No kidding. I've been writing a WinForms program that is basically a combination character progression tracking, inventory management and basic battle system for Star Wars Fantasy Flight Games.
Why? Do we not want the person-to-person experience? Hardly! We need more wetware space for strategy and thinking of creative ideas, so we need to free up the parts for calculating what number of ability vs proficiency dice and what talents may mess with the roll etc.
You still use voice (usually over discord) and DM decides what happens on the fly.
But most importantly you don't have to play with people that live close to you, makes it much easier to find someone to play with, it's why I started playing p&p rpg again after over a decade.
I could see sites like roll20 recording the sessions with metadata (at this point they said this and GM told "roll perception" and they rolled d20 + 3) as a way to train a virtual DM. Of course it would need crazy amount of games.
It also requires a lot of GM research and development, even for those games that are at least partially supported.
Roll20 also has a marginal audio chat solution. It was so bad that we chose to replace that part with Discord in my current campaign.
There is a background music solution available in roll20, but I was never impressed by it.
In my current game, we found it was better in the long run to convert everything to Discord, because it was better on the collaboration and audio chat functions, which helped the humans communicate with each other better and resulted in an overall better game. We lost the interactive mapping functions and the mostly kinda semi sorta integrated character sheets of roll20, but in the end, those were less important to us than the other collaboration functions of Discord.
There's some tips for 5e:
- set "always roll with advantage" in campaign settings to true, it removes the fiddling with the system by the player each time they want to roll, now it's just a click on a weapon or a skill, and it lets DM decide after the roll if it was advantage and change the interpretation without rerolling. You simply ignore the second roll if it was a straight test, take worse or better out of the 2 rolls if player had disadvantage/advantage
- install free chrome extension "vtt enhancement suite" - it allows for executing macros on all selected tokens at once - especially useful for starting combat with lots of enemies - you select for example 10 tokens and click on a macro that rolls initiative for each of them and adds them to the initiative tracker, you ask players to select their tokens and click the same macro - 5 seconds and combat with 16 participants is ready to go - normally I would have to roll physically 10 times, ask players and sort it manually and it would take a minute or 2
- use DM layer for hidden enemies/traps/secret doors/notes/etc - you can then easily move them to map/token layer when players roll perception
- reinforcements in a combat are easy when you put some tokens in DM layer - you can add them all to the combat tracker and the tokens on DM layer will be hidden from the players until you move them to token layer
- make players drag and drop their most used spells and items to a quick-use bar on the bottom left of the screen - it's 15 minutes to set this up on the session 0, but later it eliminates the problem "I closed my character sheet, wait 20 seconds for me to find the button to roll the attack"
- set up tokens to have hitpoints and AC bars on them (keep them hidden from players) - when you click on hitpoints bar on a token and write "-17" it will automatically decrease the hitpoints by 17 points, don't need to substract in your head every time
All of this allows me to run combat several times faster than I could possibly do physically, it adds up quickly.
Overall, in trying to reduce the complexity of the game, they gave you a finnicky spreadsheet nightmare of half-baked solutions to problems no-one had.
> Something as simple as equipping a piece of armour is an excersise in trying 5 different things
You just drag&drop it from compendium to your inventory?
Now, I haven't really played DnD (much to my sadness), but how do you win?
edit: I know the answer but I'm making a point with my question
>but how do you win?
When you've provided a fun and engaging experience for your players?
I do mean this seriously. Being a DM is about making sure the players are having fun and always giving players something interesting to do. It shouldn't be too easy or too hard. The things should have some variety, you have to be able to adapt to what your players decide to do. Just because you have some cool stuff planned doesn't mean they'll find it, you can guide them a bit, but don't funnel them and accept they may not see your cool stuff, you can always save it for another time.
I guess my point is, keeping player fun and engagement high while allowing the game to run its natural course would be what I would think of as the win condition for a dm. Finding a way to measure this and provide this seems like a fairly difficult task though.
In other words those victories need to be earned, which is a fine balancing act since dice rolls are random and bad things happen.
It can also lead to awesome situations when the dice go extremely well, everything lines up, and the players feel like legit bad-asses. It is the DM's job to create situations where that can happen!
For dungeons at least i try to go for that nethack or metroidvania vibe of slowly going deeper while your supplies and health dwindle, so each encounter needs to be considered more as they get deeper.
Above world adventures I try and let them be more free in their decisions but keep the consequences a bit higher for things.
I've yet to play any computer or console rpgs that really come close to the feeling of playing a good d&d game, even the ones that heavily use the rules. Nwn online came close, but those typically had human DMS the single player game was fun, but even on coop, still wasn't really like playing dnd without a DM player.
(A flying castle attacks the city; it's made to crash; someone has to go in there and find out what's up; cue our heros; they enter through a breach in the wall into a kind of internal moat; it's dark, you're up to your waist in water, and there's something in here with you! (Water Weirds. Damn near invisible snake-like water elementals...) Good times!)
I borrowed it from SCP series - it's a "memetic hazard" - nonmagically infects brains when it's thought of or spoken of. Mechanically players roll a Wisdom save whenever there's infection opportunity, on a fail they start shouting the meme (then others have to roll :)), on a bad fail they become possesed and run somewhere. You of course don't explain it to the players, just ask them to roll when needed :)
As they explore searching for the artifact they were hired to recover they find the whole mine is full of dead bodies calmly sitting in circles, pools of dried blood, all the books are burned and there are bodies of dwarves who were fighting each other before they died. Then there's the last room with a huge portal that starts to activate when players who got possesed run into it.
The climax is a big pvp fight between the players that try to teleport to their new god and the rest of the party trying to knock them unconscious and drag them away before the portal activates :)
Later on in the wrecked flying castle there was a room with a clockwork horror in it that I was sure would give them a hard time. They trapped it in a chest right off and ran away before it broke out.
Another time I had a classic dungeon set up for them complete with a lich. They woke up the lich... and ran away. Like all the way out of the dungeon and then halfway across the freaking continent. They just left it for someone else to deal with.
Unrelated story time...
In another campaign one player had this dwarf fighter and he had rules-lawyered and finagled to get him specialized with two axes and multiple attacks per round. Three with the right and two with the left, five attacks per round. We nicknamed him "the Blender" because he could just chop the shit out of anything. He was stupid and reckless, a follower of the great god of war VTRE (Victory Through Reckless Endangerment!) and he was awesome.
This was a Spelljammer campaign (am I the only one who loved/loves Spelljammer!? It was so much fun) and one battle in particular they were fighting a ship of mindflayers. Scary right? The Blender jumps on a loaded ballista and tells the crew to launch him at the Illithid ship. He crashes into the deck, bounces up, grabs his axes from his belt, and proceeds to straight butcher the mindflayers. I forget the details but it was one of those times that the whole table is howling, the dice are hot, and it's just unbelievable but so fun no one cares.
Great source of ideas for scary monsters/dungeons/villans/legends after reskinning from modern times to whatever setting you need.
Have a hard-nosed, Old School Revival (OSR) group who enjoys hack and slash dungeon crawls? Great, all you need is an AI who picks hard but fair challenges, describes dire and interesting settings, and keeps rewarding moments (in this case, hard-fought victories) coming frequently.
Like to tell stories instead? Seed a plot with a laid-back AI who leaves lots of room to breathe, throws lots of skill checks at crafty players, and excels at creating complex personalities and interactions via a cast of interesting NPCs.
Rimworld is an simulation/tactics/base-building game that offers more traditionally game oriented "AI Directors," but the different kinds of directors offer different prioritization over moments of peace, intense challenge, battle, etc., based on just a few settings. Even blowing out that settings list a bit more could create interesting results.
Rimworld is like 2d dwarf fortress, but way more accessible. The controls make sense, and the game teaches you as you play. Also Rimworld has about 3 and a half metric tons of mods.
I first saw RimWorld while chilling in a hotel room watching YouTube a while back, and I bought it back while it was still in early access. Tynan has done a phenomenal job making the game go from good to fantastic.
So, what "winning" looks like depends on the group and the game. Having one's character die may feel a lot like losing at one table, but just another part of the story at another, meaningful, maybe, but not a "loss" any more than passing GO and collecting $200 is losing—roll up another character, the story goes on, and we've closed a meaningful chapter in it.
At most tables, I think, the rules basically exist just to give structure and fair-feeling challenge to an extended session of play-pretend.
And come to think of it, it is not so different from ads and it is a bit scary. An AI able to provide an exciting game of DnD is going to use a lot of psychological manipulation, which is great when playing a game, much less so if you someone is pushing an agenda.
BTW, I know more about video games than tabletop RPGs but let me tell you that video games manipulate you constantly, for your pleasure. Illusion of choice, random numbers biased in your favor, the last bit of life that is worth a lot more than the rest, enemies you don't see that are slowed down, there are lots and lots of these. And these are just the tricks that make the game more enjoyable, not the dark patterns of games with in-app purchases.
Building a AI DM that can follow a scripted pre-written dungeon would still be a challenge, building an AI that can understand how the players are experiencing the gaming session to know if they want things moving faster, or they want more elaboration, or more unexpected experiences... you're really getting to the point of Ava (Ex Machina) and even beyond Data (Star Trek).
Even better, you _hide_ from your players the fact that the player's character has died (according to the dice) and instead present the possession event as the result of the other players' actions. "You cast the spell but it rebounds on the Dark Cauldron and pushes Janice's spirit through her body and into the body of the Goblin Scroll Carrier, narrowly missing the Goblin Shaman standing in front of him (or is it her?). Janice- you suddendly find yourself looking at your spasming body on the floor. You feel something sticking up your nose. You realise with horror that it's not your nose, it's the goblin caddy's nose and that thing sticking up it is his finger. Actually, it's now your finger. You're now a goblin."
Then you send the players on a quest alright- a quest to find a McGuffin that can reverse the situtation and put the PC's spirit back into her original body.
The point of course is that you don't have to kill off the character completely, so you avoid all the delay caused by the character's player having to roll a new one while the others are fighting and you avoid the awkward situation where the player ends up having to choose between the character that died and the new character (at the point where the old one can be resurrected).
I'm pretty sure that it should be perfectly possible to teach a neural net to do that.
(No, I'm taking the piss- there's no way you can teach a neural net that).
[Edit: all that goblin body swap stuff works a lot better if the PC was a paladin and if they were the character of a player who tends to play paladins and knights and whatnot. MWAHAHAHAHAH! Try teaching _that_ to a neural net.]
By spending quality time with your friends and having a lot of fun imagining an epic quest?
There seems like a vast difference between music and D&D stories/games, but just yesterday there was that RNN-sci-fi-editor experiment on here ( https://news.ycombinator.com/item?id=22427958 ) so maybe it's not as far-fetched?
In any event, being a good DM is really really fun, I don't think I'd want to let a computer muscle in on it, eh?
Those are the rules, not the game.
Turing's DM as it were
Eurisko was GOFAI, written in RLL, a frame-based language implemented in Lisp.
https://aliciapatterson.org/stories/eurisko-computer-mind-it...
As soon as Ys become Xs over time, and we come up with new Ys: we’re making progress.
NB: I'm not convinced this is the wrong thing to do, many of the things we've historically labeled "AI" are very obviously not, at least in retrospect.
Perhaps the cycle ends only when an AI can explain to us how we are getting it wrong.
D&D is not a game with rigorously defined rules and winning conditions, instead it's a means to an end, the end being: have fun with friends through a mutual fantasy. If you had a genuine post-singularity superhuman AGI, it would be silly to use it for D&D, because you could just directly instruct it, "Make my friends and I have fun through a mutual fantasy". This might be inherently dangerous because a truly superhuman AGI would probably be so effective that you and your friends would end up unable to distinguish said mutual fantasy from reality and you'd all end up blissfully running around in the resulting fantasy for the rest of your lives.
Like I say, this kind of software, pretty common in C-RPGs, is very limited when compared to a human GM, in particular because it cannot present the player with genuinely novel situations and cannot respond to the player's actions with genuinely novel reactions. In other words, the kind of "AI" (ad-hoc game logic) that runs on C-RPGs, is extremely primitive and very unsatisfying, hence the reliance of most C-RPGs on action rather than _inter_action with the players.
Since these kinds of ad-hoc "DMs" are so very limited, even a very small improvement in their ability to interact with the player(s) in surprising and imaginative manners would be a _huge_ improvement.
In the same sense, the ability to reliably generate an interesting story _procedurally_ would be a _huge_ improvement on the current situation, where the majority of procedurally generated game content is just environments with enemies, rewards etc. There are very few games that go a little bit beyond that, e.g. Dwarf Fortress that generates an entire world's mythology and so on.
So there's no need to go as far as AGI. We likely don't even have to go as far as machine learning. The kind of improvement that most gamers would really enjoy is probably within our technological capabilities right now- but there's very little in terms of efforts to make it happen. Why, I'm not sure.
A C-RPG captures D&D in the same sense that a WWII-themed video game captures the true World War II. In other words: not at all. A WWII-themed game applies WWII costumes, sound effects, etc. to an underlying game logic. Switching the WWII costumes to D&D costumes and marketing the result as D&D doesn't really make it D&D.
Except we don’t have that. AI research is mostly about taking a problem that’s just a little to difficult to actually do and working on it. A great DM AI May not need to understand what’s going on any more than a jeopardy AI, it just needs to be good enough to work.
A great many things like image classification initially look like they need AGI, but you can make something useful without it.
Edit: I personally know a good amount of people for that is a life goal.
[1] https://wiki.lesswrong.com/wiki/Wireheading
[2] https://slatestarcodex.com/2020/01/30/book-review-human-comp...
Correction: it was downplayed by Stuart Russell. Stuart took simplistic strawman examples and made them look silly by assuming perfectly calibrated common sense in the hypothetical AGI. (Based on skimming the link you posted; I haven't read Stuart's book.)
On the flip side, this is what people said. Forget Chess - Go is the one. Forget Go - D&D is the one.
If someone can make even a decent AI DM, I'd say turn that one loose on NP-complete for a followup softball question.
Being a DM - or a player - can be very challenging indeed.
the thing about 70s style D&D is the liberal arts involvement as role playing is really thespianism
the costumes and the soapboxing as people play out thier characters was really something
Original 70's DnD started as players vs a killer dungeon built by the DM Aka Gary Gygax
Ironically it was those of us who crossed over from traditional wargames who where more into the role playing your character.
You can see more recent players in pathfinder and dnd 5th wanting to win "the game", the full on RP style of the Mighty Nine of The Glass Cannon is quite different are actors or semi pro ones AND THEY ARE SO WELL BEHAVED""
The more structured and rules based the game is, the easier it is to create a dungeon master - the more free form and "on the fly" the game is the harder the task. So while creating a dungeon master may be "easy" if the rules and story are heavily enumerated, it likely won't be very fun as it doesn't capture the free-formedness that makes D&D so fun.
D&D at least has well defined combat rules. I've ran games that use cinematic combat rules, where the books have suggestions and some math, but a lot of things are left up in the air with the general guidance of "whatever is awesome."
So there aren't rules for shooting grenades out of mid air, but it made for an awesome climax to the story as the team raced onto the escape ship and the hanger bay door was slamming down, so of course I allowed it.
I was playing one game where we the players objected to how the military was prepping for an AI invasion so we rebelled, next session the DM brought in a giant map and it turned into a base building tower defense session.
D&D can go off the rails, other systems can go off the rails then continue until things are off planet!