AI-GMs


Pathfinder Second Edition General Discussion

151 to 181 of 181 << first < prev | 1 | 2 | 3 | 4 | next > last >>
Liberty's Edge

1 person marked this as a favorite.

AIs do the boring job humans could do, but faster. So, they can be used as help with good value. But they will not be able to provide the human touch (aka those things not boring).

Sczarni

Pathfinder Lost Omens, Rulebook Subscriber

Yall realize the line between machine AI and being able to differentiate it from a human is becoming blurrier and blurrier, right? In 20 years... we won't be able to tell the difference.. and AI will surpass every Turing test we have.


Verzen wrote:
Yall realize the line between machine AI and being able to differentiate it from a human is becoming blurrier and blurrier, right? In 20 years... we won't be able to tell the difference.. and AI will surpass every Turing test we have.

There's a big difference between "bot or not?" for a Turing test as if dealing with a stranger and an active GM-friend developing a long-term relationship w/ both players and their PCs over the course of an extended campaign narrative incorporating backstories, tangents, and wrenches.


9 people marked this as a favorite.

ChatGPT is definitely not an AGI. It's a chatbot and chatbots have been fooling people for years so that's hardly a milestone. Besides, with the amount of AI-generated media already flooding what they pull to generate their models, they may end up essentially poisoning themselves with compounding errors and becoming functionally unusable without a person holding their hands the entire way.

AI-generated stuff is just a shallow fad. It's doubtful we'll see an AGI in our lifetimes much less one I'd think of as a good GM.


6 people marked this as a favorite.

There's nothing remotely inevitable about AI ever improving past the average person to become "better than any human GM", because "better" turns it into an entirely separate problem that nobody has begun to solve. If you want "average", you can dump in anything at least passable as training data, but there's precious little training data for exceptional GMing. Even exceptional writing in general, pulling from centuries of close-enough-to-modern-speech writings, is really hard to pin down.


5 people marked this as a favorite.

Nevermind that current AI generators are garbage at long-term planning, understanding context, understanding theme, or so many other nuances that make a good GM.

There's plenty of tools and books out there to generate contextless combat encounters and dungeons if you really need that itch scratched. What AIs are doing is really nothing new or remarkable. In fact, I'd argue they're kind of bad at it.


1 person marked this as a favorite.
Castilliano wrote:
Verzen wrote:
Yall realize the line between machine AI and being able to differentiate it from a human is becoming blurrier and blurrier, right? In 20 years... we won't be able to tell the difference.. and AI will surpass every Turing test we have.
There's a big difference between "bot or not?" for a Turing test as if dealing with a stranger and an active GM-friend developing a long-term relationship w/ both players and their PCs over the course of an extended campaign narrative incorporating backstories, tangents, and wrenches.

Long term relationships are not exactly useful to GM. I play a lot of PFS with strangers and it works fine.

Also, backstories, tangents and wrenches are not a given, even with a human GM.

In my opinion, AIs will eventually be able to replace a human GM entirely (maybe not the best human GM, but I've played with a lot of mediocre human GMs and the sessions were still nice). The biggest issue is that no AI can understand and apply game rules. So, the AIs will need rule-specific subAI to handle, at least, combat. And that's, in my opinion, the biggest issue as of now.

But for rule light games, I expect AI GMs to arrive quickly. We can see it with the link I've given earlier: There are already people trying to achieve that. I'm pretty sure they'll get some results.


1 person marked this as a favorite.

So... what you want is just a procedurally generated dungeon crawl with a half-way decent character creator? I feel like there's already plenty of non-AI versions of that on the market that are more robust and have higher quality content.

Also, all of this is predicated on the assumption that decent AI generators won't be licensed to hell and back forcing you to pay out the nose for access to a service that your buddy at the FLGS would do for a couple of root beers and a pepperoni pizza.

Radiant Oath

Master Han Del of the Web wrote:


Also, all of this is predicated on the assumption that decent AI generators won't be licensed to hell and back forcing you to pay out the nose for access to a service that your buddy at the FLGS would do for a couple of root beers and a pepperoni pizza.

The limit is not money, but time.


SuperBidi wrote:
Castilliano wrote:
Verzen wrote:
Yall realize the line between machine AI and being able to differentiate it from a human is becoming blurrier and blurrier, right? In 20 years... we won't be able to tell the difference.. and AI will surpass every Turing test we have.
There's a big difference between "bot or not?" for a Turing test as if dealing with a stranger and an active GM-friend developing a long-term relationship w/ both players and their PCs over the course of an extended campaign narrative incorporating backstories, tangents, and wrenches.

Long term relationships are not exactly useful to GM. I play a lot of PFS with strangers and it works fine.

Also, backstories, tangents and wrenches are not a given, even with a human GM.

In my opinion, AIs will eventually be able to replace a human GM entirely (maybe not the best human GM, but I've played with a lot of mediocre human GMs and the sessions were still nice). The biggest issue is that no AI can understand and apply game rules. So, the AIs will need rule-specific subAI to handle, at least, combat. And that's, in my opinion, the biggest issue as of now.

But for rule light games, I expect AI GMs to arrive quickly. We can see it with the link I've given earlier: There are already people trying to achieve that. I'm pretty sure they'll get some results.

We aren't talking about "works fine", we're talking about excellence.

I'm pretty sure AIs will be able to "work fine" for brief runs with narrow boundaries and few if little ties to the PCs, like most PFS scenarios.

Yes, being able to integrate backstories, run with or smoothly clip tangents, and catch wrenches before they muck up the machinery are not givens for a human, but again we're talking about excellence here. A good GM can do these things while a top-tier AI cannot. We're talking the difference between a superior novel or poem (whatever that means to each individual, or table in this case) & a choose-your-own-adventure, with PFS being fanfic/slam poetry somewhere in-between that an AI can kinda handle, yet it's still lacking.


3 people marked this as a favorite.
Castilliano wrote:
We aren't talking about "works fine", we're talking about excellence.

I disagree. If I had to play only with excellent GMs, I'd have not played much. If everyone only plays with excellent GMs, no one plays as excellent GMs are not made in a day.

Also, as Ace says, the main asset of AI is availability and lack of time to prep. That's a major asset, one that no human can beat. Considering that AIs have to fight humans on human ground is a mistake, AIs will fight humans on many grounds and some of them will see the AIs thrive.

In my opinion, the only question is when (and, subsidiary question, how much). When it will be time, and if the development of such AI is not too expensive considering the niche hobby that we have, we will have competitive AIs.

In my opinion, there will be 2 types of GMing for 2 types of RPGs in the future. Those who love stories, backstories, complex stories, NPC interactions will keep a human GM. Those who prefer the more mechanical aspects, who don't care much about backstories or complex NPC interaction but want to keep the fun of playing with friends will move to AI GMing.


3 people marked this as a favorite.
AceofMoxen wrote:
Master Han Del of the Web wrote:


Also, all of this is predicated on the assumption that decent AI generators won't be licensed to hell and back forcing you to pay out the nose for access to a service that your buddy at the FLGS would do for a couple of root beers and a pepperoni pizza.
The limit is not money, but time.

Any discussion that does not on some level include money in its considerations is short-sighted. Cost is one of the major contributing factors as to whether there will be a version of the product that can be adopted by the public in a meaningful way and iterated upon based on their wants and desires as opposed to remaining say... a method for Disney to pay fewer writers and artists on the next multi-mega-crossover marvel superhero event featuring Chadwick Boseman's ghoulishly digitally resurrected image.

(Man, it's a sad day when the communist has to remind people to think in terms of capitalism)


5 people marked this as a favorite.
SuperBidi wrote:
Castilliano wrote:
We aren't talking about "works fine", we're talking about excellence.

I disagree. If I had to play only with excellent GMs, I'd have not played much. If everyone only plays with excellent GMs, no one plays as excellent GMs are not made in a day.

Also, as Ace says, the main asset of AI is availability and lack of time to prep. That's a major asset, one that no human can beat. Considering that AIs have to fight humans on human ground is a mistake, AIs will fight humans on many grounds and some of them will see the AIs thrive.

In my opinion, the only question is when (and, subsidiary question, how much). When it will be time, and if the development of such AI is not too expensive considering the niche hobby that we have, we will have competitive AIs.

In my opinion, there will be 2 types of GMing for 2 types of RPGs in the future. Those who love stories, backstories, complex stories, NPC interactions will keep a human GM. Those who prefer the more mechanical aspects, who don't care much about backstories or complex NPC interaction but want to keep the fun of playing with friends will move to AI GMing.

The rise of video games promised literally everything you think AI will do for TTRPGs but we still have TTRPGs run by humans. It's almost like playing with other humans has some benefit a chatbot cannot provide.


SuperBidi wrote:
Castilliano wrote:
We aren't talking about "works fine", we're talking about excellence.

I disagree. If I had to play only with excellent GMs, I'd have not played much. If everyone only plays with excellent GMs, no one plays as excellent GMs are not made in a day.

Also, as Ace says, the main asset of AI is availability and lack of time to prep. That's a major asset, one that no human can beat. Considering that AIs have to fight humans on human ground is a mistake, AIs will fight humans on many grounds and some of them will see the AIs thrive.

In my opinion, the only question is when (and, subsidiary question, how much). When it will be time, and if the development of such AI is not too expensive considering the niche hobby that we have, we will have competitive AIs.

In my opinion, there will be 2 types of GMing for 2 types of RPGs in the future. Those who love stories, backstories, complex stories, NPC interactions will keep a human GM. Those who prefer the more mechanical aspects, who don't care much about backstories or complex NPC interaction but want to keep the fun of playing with friends will move to AI GMing.

Don't mistake me. I'll settle for a "choose your own adventure" when deprived of a better novel. Heck, I bested Baldur's Gate with maybe every class, most often solo, simply to get my RPG fix. But the question was if AI's could emulate a good GM, wasn't it?

IMO they can't, though yes, as you've noted, there's another level of GMing they can fulfill. I'd say with human scripting, AIs already can...if the players are willing to navigate within such mechanical parameters, stick to a minimalist form of RAW. Those parameters kind ruin the creative aspects to me, though yes, a table full of gregarious players can overcome such limitations. I've played bad RPGs simply for the company, but that's not GMing, that's having fun peers who could probably riff off on their own or from other inspirations, like improvising a sequel to Princess Bride. And it's hard to imagine an AI hearing player interplay and doing anything substantial with it or contributing which is something even mediocre GMs can do.


3 people marked this as a favorite.
Master Han Del of the Web wrote:
The rise of video games promised literally everything you think AI will do for TTRPGs but we still have TTRPGs run by humans. It's almost like playing with other humans has some benefit a chatbot cannot provide.

Video games are a substitude for TTRPGs. They haven't replaced them entirely, but they still have a niche alongside TTRPGs (well, the niche is TTRPG actually despite being the earlier type of games).

AIs will go further by enabling more interactions than video games when it comes to the story, the descriptions and the ability to interact with the environment.


1 person marked this as a favorite.
SuperBidi wrote:
Master Han Del of the Web wrote:
The rise of video games promised literally everything you think AI will do for TTRPGs but we still have TTRPGs run by humans. It's almost like playing with other humans has some benefit a chatbot cannot provide.

Video games are a substitude for TTRPGs. They haven't replaced them entirely, but they still have a niche alongside TTRPGs (well, the niche is TTRPG actually despite being the earlier type of games).

AIs will go further by enabling more interactions than video games when it comes to the story, the descriptions and the ability to interact with the environment.

So... they provide the benefits over video games that you said earlier that a specialized human GM would provide over an AI GM? It really sounds like you're splitting hairs here to try and find a niche for them.

Liberty's Edge

2 people marked this as a favorite.

AIs will be great for GMs who love complicated and subtle stories but lack the inclination and expertise to build all the hazards, maps, monster / NPC stats...

I know I would love having one that would do this tedious heavy lifting for me while I focus on the plot and the personalities.


2 people marked this as a favorite.

In theory assuming a powerful enough AI then:

AI controlled procedurally generated sandbox style and improv style is better for VTT and Video games.

Human scripted AI controlled games are better for accessibility given it allows people who normally could not play to play. It is also one step above a VTT like foundry by letting the AI handle all the software while the humans do the actual story.

AI scripted Human controlled games are better for IRL improv with the AI acting as a neutral party.

Classic video games are better for experiencing a set story or game as intended by the creator.

Classic TTRPGs are better for offline play and just casual fun.

*******************

In any case, people will not stop doing physical version just because a digital version of it exists. Will it become more rare? Sure. Will it be eliminated? Nope.


1 person marked this as a favorite.

God this whole conversation is grim


1 person marked this as a favorite.
Master Han Del of the Web wrote:


So... they provide the benefits over video games that you said earlier that a specialized human GM would provide over an AI GM? It really sounds like you're splitting hairs here to try and find a niche for them.

The benefit is that you play with your friends. You can play video games with your friends but video games rarely encourage deep interactions between characters like inter PC roleplay.

And as far as I'm concerned, human interactions is the selling point of TTRPGs over video games.


I wouldn't pay for it. I might use some program for fun to see how it works.

Since Tabletop RPGs are as much a social activity as they are a game, unlikely I'd use an AI GM over a human one. People who like to GM enjoy the creativity and storytelling. If I wanted a GM that generated an AI-like experience, I'd play a video game.

I think AI will have more of an effect on video games than TV, movies, and things like Tabletop RPGs where people prefer the creativity of humans.

Maybe big corporations would put it out for consumption by the masses, but I think most humans would still prefer to interact with other humans for things like movies, TV shows, and Tabletop RPGs with maybe some AI tools to supplement.

As a GM I'd love an AI tool for map drawing or rendering an enemy or doing a voice when you feed it dialogue. That would be kind of neat.


People have already experimented with AI DMery and it tends to spit out a ton of narrative content - far, far, far more than you would expect or want a DM to when you do something as simple as say "I attack the ogre".

Furthermore, when people are white room theory crafting, navel gazing, or otherwise pontificating about AI they forget:

People want to play with people. The entire concept of an AI DM, or players for that matter, is utterly reductive and backwards.


5 people marked this as a favorite.

For a long time I've struggled with "naming NPCs" and I freely admit to giving ChatGPT prompts like "give me 30 potential names for [character I describe]" and then picking or mixing/matching.

This is a lot quicker than my former process of "agonizing for a while until inspiration hits."


1 person marked this as a favorite.

Honestly, if you were to have a fully AI GM, at that point I'd feel like you were playing a video game. Which isn't necessarily a bad thing, I love some good co-op games with friends, but it is a very different experience compared to a tabletop game. At that point, rather than paying for an AI to fumble around with tabletop rules/simulation, I'd rather just pay for a video game that selectively uses AI to complement certain features. (I'm actually playing one now that uses simple AI generation for a world of NPCs that develop/interact/grow each month in-game time. While very interesting - there are a number of noticeable issues with an AI handling most NPCs, the most obvious of which is how similar all the NPCs end up being.)

Themetricsystem wrote:

Oh boy, you might be surprised to hear about this but there are MASSIVE Korean and Japanese companies like Hololive that already ARE beginning to roll this kind of thing out with AI-driven chatbot Vtuber streamers and content creators that are raking in money by the truckload.

It is way more affordable to pay a handful of developers to maintain AI and hire artists to create/customize/model/decorate 3d avatars and just duplicate those efforts n+1 times to create their own fleet of AI-generated influencers/actors than it is to seek out, train, coach, and pay actual talented, relatable and attractive entertainers or content creators.

Guessing you're not part of the vtuber fandom, as this is a (depressingly) frequent misunderstanding of how vtubers work. The vast majority of vtubers are actual people using Live2D avatars. Hololive does not have any "AI-driven chatbot Vtuber streamers and content creators." The only actual AI vtuber I can think of is Neuro-sama, whose popularity is partially because on how awkward, random, and inconsistent the AI component is when trying to interact with actual people (along with the developer's constant struggles to keep them in-line with socially acceptable behavior and censoring outputs to avoid getting banned by the streaming platform again.)

Radiant Oath

Master Han Del of the Web wrote:
AceofMoxen wrote:
Master Han Del of the Web wrote:


Also, all of this is predicated on the assumption that decent AI generators won't be licensed to hell and back forcing you to pay out the nose for access to a service that your buddy at the FLGS would do for a couple of root beers and a pepperoni pizza.
The limit is not money, but time.

Any discussion that does not on some level include money in its considerations is short-sighted. Cost is one of the major contributing factors as to whether there will be a version of the product that can be adopted by the public in a meaningful way and iterated upon based on their wants and desires as opposed to remaining say... a method for Disney to pay fewer writers and artists on the next multi-mega-crossover marvel superhero event featuring Chadwick Boseman's ghoulishly digitally resurrected image.

(Man, it's a sad day when the communist has to remind people to think in terms of capitalism)

I'm an economist. Things are maximized to the limit of their scarcest input, relative to the size of that input. If your pie recipe takes 2 apples and one pound of bread, the limit is apples divided by 2 or pounds of bread. In a market system, you can trade money for the scarcest input.

However, while you may trade your time for others' money, there are limits on trading your money for others' time. The market may exist for ten Douglas Adams books a year, but he famously took forever to write a book, and only got them out at a much slower rate. Ironically, in fact, his increasing pay tended to lower his output, with new books coming out as he was about to run out of cash.

The money will be there for some segment of the population, but the time will not be for many people. Further, AI use is already nearly free, and technology costs tend to come down. It's possible that, like social media or Uber, the Monetary costs could slowly rise instead, but I doubt that they will lock as many out as the time constraint.


AceofMoxen wrote:
Master Han Del of the Web wrote:
AceofMoxen wrote:
Master Han Del of the Web wrote:


Also, all of this is predicated on the assumption that decent AI generators won't be licensed to hell and back forcing you to pay out the nose for access to a service that your buddy at the FLGS would do for a couple of root beers and a pepperoni pizza.
The limit is not money, but time.

Any discussion that does not on some level include money in its considerations is short-sighted. Cost is one of the major contributing factors as to whether there will be a version of the product that can be adopted by the public in a meaningful way and iterated upon based on their wants and desires as opposed to remaining say... a method for Disney to pay fewer writers and artists on the next multi-mega-crossover marvel superhero event featuring Chadwick Boseman's ghoulishly digitally resurrected image.

(Man, it's a sad day when the communist has to remind people to think in terms of capitalism)

I'm an economist. Things are maximized to the limit of their scarcest input, relative to the size of that input. If your pie recipe takes 2 apples and one pound of bread, the limit is apples divided by 2 or pounds of bread. In a market system, you can trade money for the scarcest input.

However, while you may trade your time for others' money, there are limits on trading your money for others' time. The market may exist for ten Douglas Adams books a year, but he famously took forever to write a book, and only got them out at a much slower rate. Ironically, in fact, his increasing pay tended to lower his output, with new books coming out as he was about to run out of cash.

The money will be there for some segment of the population, but the time will not be for many people. Further, AI use is already nearly free, and technology costs tend to come down. It's possible that, like social media or Uber, the Monetary costs could slowly rise instead, but I doubt that they will lock as many out as the time...

Just like with how making a high resolution good looking video game has become much easier. The cost and issue now being in making sure it runs correctly.


7 people marked this as a favorite.
Themetricsystem wrote:
Ashbourne wrote:
So what comes after AI DMs, AI Players? With both WotC could replace Critical Role with an all AI cast. With enough AI twitter and youtube bots commenting and liking it's bound to be a big hit.

Oh boy, you might be surprised to hear about this but there are MASSIVE Korean and Japanese companies like Hololive that already ARE beginning to roll this kind of thing out with AI-driven chatbot Vtuber streamers and content creators that are raking in money by the truckload.

It is way more affordable to pay a handful of developers to maintain AI and hire artists to create/customize/model/decorate 3d avatars and just duplicate those efforts n+1 times to create their own fleet of AI-generated influencers/actors than it is to seek out, train, coach, and pay actual talented, relatable and attractive entertainers or content creators.

I am a vtuber and there is exactly one (1) AI/chatbot vtuber, Neuro-Sama, and they're entirely a gimmick/meme.

Vtubers are just people who stream using an avatar animated by live2d.


2 people marked this as a favorite.

I think that long term, AI tools will be like the calculator. If you are a skilled and talented DnD creator then the tools will enhance your output and make you even more creative/productive.

The downside is there will be a lot more mediocre stuff released into the wild. AI will never be super creative without the initial input/spark from the human who has the artistic vision.

I have actually built an AI DnD character creator tool with my brother, because we hated how disempowering most AI tools are (very simple input, boring output, hardly any control, just farts out 'something')

Our idea is to provide inspiration at every step of the character building process, and works more as an inspiration tool rather then just outputting the final product. You get to be involved in the step by step process of generating a complex and interesting character.

As well, it also offers creating characters by building their lifepath in a interactive story that outputs your character at the end.

Give it a whirl! https://charactercomposer.com/


Pathfinder Roleplaying Game Superscriber

AI GM...

That's basically a procedurally generated MMO or Skyrim Mod.

Those can be OK... but they're not as engaging.

I think we'll get "AI GMs" that people actively look for rather than use out of desperation at the same point in time when you go out and look for an "AI Friend" to hang out with at the local pub or an "AI Romantic Partner" that you're going for on purpose rather than desperation...

The stuff just has to get a LOT better.

For me, a GM and the other players are all part of my social experience. Not 'content delivery services'. But then that's why I don't go in for paid games.

Given that paid GMs is a thing, AI GMs might be as well sooner than I personally would want to use one.


arcady wrote:

When you pay a GM the contract isn't 'social', it's financial. So you expect a certain level of product and you're not there to bond with each other. AI will eventually deliver a better game experience than any human GM can, and that will hurt paid GMing.

It is also performance and creative. I can see Ai GMing taking over from the most supercifical levels, but it is a long way from being able to handle the other aspects any half decent GM has to for a good game.

What it could do is replace paid GMs who run adventures almost verbatim and don't adjust anything for players, but that is bottom of the barrel stuff.

I am not a paid GM, but I:

- make visual assets (maps, images, animated effects, tokens)
- collect and organise sound effects, ambiences and music libraries (changing music play lists and tone for locations and also to keep my players feeling the right emotions, which can often change mid session)
- adjust events and progression over the campaign based on player and character interests and monitor what is resonating with players
- adjust events and progression within a session based on current player energy levels
- learn from how players respond from all of the above and use it to inform how I go about things in the future

I would imagine that most successful paid GMs do that at the very least.


Pathfinder Roleplaying Game Superscriber
The Gleeful Grognard wrote:
arcady wrote:

When you pay a GM the contract isn't 'social', it's financial. So you expect a certain level of product and you're not there to bond with each other. AI will eventually deliver a better game experience than any human GM can, and that will hurt paid GMing.

It is also performance and creative. I can see Ai GMing taking over from the most supercifical levels, but it is a long way from being able to handle the other aspects any half decent GM has to for a good game.

What it could do is replace paid GMs who run adventures almost verbatim and don't adjust anything for players, but that is bottom of the barrel stuff.

Is that a quote from me from somewhere else? I don't recall it.

But yeah. AI GMs would hurt paid GMing.

But frankly MMOs already deliver a better "gaming" experience than any tRPG can. But they deliver a vastly inferior social experience.

I play tRPGs for the social experience. I'll play a cRPG for the gamism experience.

151 to 181 of 181 << first < prev | 1 | 2 | 3 | 4 | next > last >>
Community / Forums / Pathfinder / Pathfinder Second Edition / General Discussion / AI-GMs All Messageboards

Want to post a reply? Sign in.