D&D5e: AI Project - Page 2 - Myth-Weavers


GM Workshop

A community-created and maintained place for Game Masters of all systems to bounce ideas around. It's a place for inspiration and sharing tips.


D&D5e: AI Project

   
That is how your typical game AI will work.

But I think my confusion is this - what niche is this actually filling? You're not writing actual program code for a computer game, yet it's more complicated and less user-friendly than just writing basic strategies because instead of just giving the creature's general aims you've gone right down to a pretty detailed "if X and Y then Z" level.

Well basically I am a software programmer. And I am approaching it from that angle. Instead of doing the legwork and manually figuring out the AI for each monster what you want to do is essentially create an AI generating program that you can than plug monsters into and get AI out. The entire thing should ideally just be a single big input and a giant decision tree.

If done right such a program could be made to produce either a strategy guide like the one he mentioned or alternatively be expanded to take direct situational input as well and just simple commands like "In this situation the monster will attack. Thank you for asking."


The fun part though is that if you took this further you could actually plug it into an actual learning AI and have it develop the decision rules for you in order to eventually evolve the ultimate unbeatable combat AI. But I won't pretend I have the resources or inclination to do so at this time.

Quote:
Originally Posted by PPQ_Purple View Post
The fun part though is that if you took this further you could actually plug it into an actual learning AI and have it develop the decision rules for you in order to eventually evolve the ultimate unbeatable combat AI. But I won't pretend I have the resources or inclination to do so at this time.
He said earlier he didn't want to deal with complexity, though.

What I'm having trouble with deciphering is the why. Not, why are you doing this, for fun that's obvious. But Why is this better than a human? Faster and easily reproduced does not a fun game make.

Global Variables like: Move, Seek, Attack, Flee, Kill and whatnot will simplify lots of things. But that will make very cardboard cutout and not particularly interesting creatures.

How does the AI know when to ambush the player? How does the AI react to players Rock to Mudding walls and disintegrating the dungeon? How does the AI know what spells it prepped that day? Does it decide them on the fly(that's cheating)? How does the AI distinguish between an intelligence of 3 Wolf and an intelligence of 10 Warg? Does the AI communicate with the player in any meaningful way?

I'm not trying to discourage you from doing this. Just pointing out some problems I've run into in thought experiments I've run on the same idea. The AI falls apart when you consider things from the user's perspective. Nothing prevents me from talking to my GM for a few seconds at a table, or in the OOC thread, about using a control spell like wall of iron to reshape the dungeon/environment. Ostensibly I can make a tower in a few minutes with the right spells prepared when dealing with a human arbiter. But the AI?

Think of it like this, chess AI's have billions of different moves. That's dealing with 32 pieces on 64 squares each piece with a very specific(but nigh identical function). By the time you have a 20*5 room with 3 different monster types and 5 players you're looking at quintillions of different possibilities.

More power to you if you manage to make an AI that can handle that, you'll solve the Travelling Salesman Problem. I for one would bow to our new D&D 5e AI overlord.

I feel like a lot of you are asking questions that I answered early on in my second post. I can't explain it better than I did there. Especially, if it feels like you are ignoring some of what I wrote there. All I can do is rewrite it, or copy and paste into a new post, but I don't have any inclination to do that since you may as well just ignore what I've written then, too. I can only assume that you didn't read the second post, or didn't read is closely enough.

I think you guys are reading too much into the Artificial Intelligence title. It's not meant to be that complex. It's a bit of a misnomer, or rather a complete one. All MAI is, is a simple system that tells you what monsters would do in battle. Just like the system in place in the D&D Adventures Board Game. If you don't know what that is, I suggest you look it up. It might answer all of your questions.

As I've said before, MAI isn't mean to replace a DM on her own. Read my second post more closely, where I go into detail about this. t.

Monsters in 5e already have prepared spell lists. MAI was never intended to recreate monsters. Just tactics for monsters. I already answered about different monsters with different intelligence. It's largely not necessary. A creature with more intelligence would probably have more Tactics and Tactics that are more effective than a less intelligent creature. That may not even be the case, if it's a creature with natural combat instincts. Now, I'm repeating myself.

Fred, you asked what use this kind of system can have. I detailed several uses I hope to use the system for in my second post in the topic.

As for the framework that I've been working, I decided to continue with my original idea for a little bit longer. The purpose of working on developing the Tactics for a single specific Monsters, is so we can see what kind of Tactics a single creature will need. After it's finished, and polished, we can then reverse engineer it to make a guide that can be used to apply Tactics to any monster you want to try and plug into the equation. I imagine that would be incomplete still, but it will be a good place to star

Quote:
Originally Posted by Black_Valor View Post
He said earlier he didn't want to deal with complexity, though.
Don't worry about that last bit. I said it would be cool, not legally sane to do it.

Quote:
Originally Posted by The Stranger View Post
All MAI is, is a simple system that tells you what monsters would do in battle.
Well, except that already exists - many monsters have a "tactics" entry, some of which are vague but some of which are much more specific. See the Balor, for instance. Now, those aren't very detailed or very good, but I can read them in a few seconds rather than having to crack out the Manual of Complicated Rules for Running Monsters. Given that most monsters will last like two rounds, that's really all I need.
Quote:
Originally Posted by The Stranger View Post
As I've said before, MAI isn't mean to replace a DM on her own. Read my second post more closely, where I go into detail about this. t.
...
Fred, you asked what use this kind of system can have. I detailed several uses I hope to use the system for in my second post in the topic.
Replacing a DM seems to be like the only sensible use case for it (DMless D&D), though that by that point you really are pretty much just writing a computer game. You also list "streamlining" as a use case, but as I said in my reply to that I personally don't see that it does anything of the sort. The examples you've given are anything but. I can see perhaps a nice middle ground in terms of tactics which are a bit better than the Balor's but not so convoluted as those, perhaps; personally I think I'd rather just run the monsters myself, but perhaps for the busy DM it would be useful to have a fallback.

Your only other use case is the "fairness" one, which you even say yourself is not such a big deal. In a typical DMed game, you really can't mandate DM fairness without essentially bypassing the DM completely; if the DM must abide by certain rules, then we're almost back to DMless again. A DM could always just choose to ignore the AI, and even if they didn't they could always just choose more powerful monsters, etc. And if you're mandating that the DM follow your rules, you quickly get bogged down in the problems many rules-heavy systems have with funky edge cases and loopholes as far as the letter of the law goes. Forget some unusual scenario and now the DM's not allowed to adapt to it? You end up with the same sort of cheesy exploits that you get in PC games (like casting at bad guys who are out of vision range but which you know by metagaming are there, etc) - because that's what you've basically produced, a DMless PC game.

So... some guidelines, might be handy, might be fun. I often jot down guidelines for myself if I'm creating an NPC foe, if only to remind myself of the options that character has when I'm running them. My point about use case though is not that you don't necessarily have one at all but that different use cases call for different approaches, and this seems too (needlessly) complicated for being guidelines but is not detailed enough to be a codified piece of software (which I'm pretty sure you'd not want anyway).

The spirit of my inquiry is to merely address why the AI is beneficial or if it is even possible. In order to make a program that you suggest you'd have to first figure out if P=NP or P!=NP. Because no matter what you're trying to do with this program, it will need to make a decision based on literally infinite variables.

Now if you wanted to do something truly awesome for D&D. Here's my two cents.

Instead of trying to make a monster AI, create an environment generator. Firstly, there are already programs capable of dynamically generating tilesets that contain both monsters, loot, and resolvable logic puzzles. These programs would take 90% of the work out of GMing. While still leaving the GM to create the story, the GM only needs to click a few toggles and create an environment to play the story in. In fact, these tools can assist in the creation of a story because it allows the GM to use placer memory instead of having to imagine an environment. Thus, the restriction created by the tool allows creativity to be focused on making an immersive dialogue with the player, rather than creating a challenging tileset and trying to work a story into it.

I really don't see a use for this. The whole point of a TTRPG is for it to be a customized imaginative story. Part of that is descriptive battles which I do both in person and over voice.

I suppose you lose that in PbP, making battles long tedious affairs, but I see it more as a reaction to the medium of play as opposed to the genre needing it.

I wish you luck finding someone to do all this work for you..

The challenge your question presents is it goes against the nature of D&D. I'm not saying you can't or shouldn't work to find a way to make it not reliant on a DM if that's what you want, only that the system is built around the idea that it has a DM to roleplay all the NPCs (including monsters) and be the arbiter. My best advice would be to look for an indy RPG system or a collaborative board game with a fantasy element to it to meet your group's needs.

If you do want to pursue a more DM-less system, here's a few options I thought of:

-Don't focus on making rules for every monster. Pick 10-20, and come up with rules for those. A beast with low intelligence may go after the biggest threat (most damage last turn) it can reach, a crazy monster might take an action based off a random d6 roll, a monster with a particularly favored meal or enemy may be predisposed to attack a character based on their race or class.

-Have players take turns making the dungeon. To a degree, someone has to set the stage, and by taking turns, you'll only have one PC at any one time have an idea of what's in the next door.


As for DM/PCs in general, what I've found is that it can work if the DM plays a character who is weaker than the characters (a level or two lower) and/or is very temporary. They get the Dwarf guide through the mountain pass, but then he leaves. Later on, they meet a pair of Paladins hunting the same necromancer. Once that's done they go their separate ways, and the PCs go on a new adventure guarding the Prince who can defend himself (though not as well as he thinks).







 

Powered by vBulletin® Version 3.8.8
Copyright ©2000 - 2019, vBulletin Solutions, Inc.
User Alert System provided by Advanced User Tagging (Lite) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Last Database Backup 2019-02-20 09:00:08am local time
Myth-Weavers Status