Are ChatGPT, Gemini, and other AI good for the TTRPG hobby?

How do you feel about the use of AI in the tabletop roleplaying games that you play and run?

Transcript

When the algorythm-driven image generators hit the scene, it didn't take very long for the community around tabletop role-playing games for the most part to disavow them. Though some of us might have been a half step slower than others.However, I've noticed an uptick or several upticks in the talk around large language models, things that are used in chat bots, such as chat GPT, for use in tabletop role-playing games. And while the image generators were pointed out, rightfully so, for the horrible impact they have on the commercial side of gaming, because they have not only the stealing from artists, but they also have the capability to replace artists. But I think the threat, maybe that's a strong word, but in some ways I think it's accurate, the threat of the large language models is against the very content of the hobby, which to me is more central than the art, because ultimately we're playing a tabletop role-playing game, we're not playing the art, we're playing the content. And if the creation of that content gets shifted to these large language model algorithms, programs, or whatever you want to call them, at a certain point, is it even the same hobby anymore? What are we doing?

I have been wanting to talk about this for a while, but I haven't, and I don't think I have now, even at this moment, figured out the best way to get the point across, talk about what I want to talk about. I'm not trying to tell anybody you're playing the game wrong, okay? I think that's a response I see a lot when I've seen these discussions about the use of things like chat GPT in games is, "Don't tell me I'm playing the game wrong." I thought D&D was about having fun. It's not about that. It's fair for people to want to preserve the humanity in the games. And what these games are really about is that creativity that comes from our humanity.

Are we offloading our creativity to AI tools?

You can look at how these tools, or through these tools, if you even want to call them tools anymore, and I'm not sure at what point it stops becoming a tool, we're offloading our creativity, the thing about us that we're putting into the game, to these LLMs, to these apps, to these chat bots. At a certain point, it stops becoming our thing, and it becomes their thing.

Maybe the best analogy that I've come up with. If you're in the corporate world, there are managers and then there are doers. And I don't mean to say that managers don't do anything, but the folks who are creating stuff, and then you have the people who are managing that creation. Programmers are programming. Graphic designers are creating graphics. You know, videographers are creating videos. Actors are acting. People are doing things. You have these people who are doing things, and then you have these folks who are managing the doing of things. They may approve something or decline something. They may have final say on something. They may delegate something, but ultimately, they are no longer in the field of creation. They are in the field of management and not really to denigrate managers. But I think at the end of the day, most of us, I know I do, value the creation more than the management. Sure, there are skills involved with managing, and there are some great managers and some terrible managers, just like there are great creators and not so great creators. But ultimately, the creation is the thing. If you take away all the creators and you just have all the managers, nothing's going to get done.

If you have a bunch of creators and no managers, it's probably going to be a very inefficient process, but probably some stuff will get made. I'm not to say that there is no value to being managers. There certainly is, but I think we'd rather have if we had to choose the creation over the management.

Are AI tools transforming us from TTRPG creators to TTRPG managers?

What does this mean in terms of tabletop role playing games and large language models? Well, I think with the large language models, we're offloading more and more of this creation to them, and we're taking on the management role. We're writing prompts. Essentially, we're delegating. Maybe we're approving at some point or declining, saying, "No, do it again." But we're not creating anymore.

Anytime somebody says this, you get people coming out of the woodwork with all the exceptions. "No, no, no. For me, I'm doing it in all these ways that it's totally appropriate and totally responsible, and therefore, you're wrong and I'm right, and this is just a tool, and you're a Luddite, and ha, ha." Okay, there are always going to be exceptions. I have no doubt. I believe you. If you say, "I've used these tools in a responsible manner, and I'm not managing, I'm using as part of my creative workflow," fine. But I think we've seen already, and we will continue to see a lot of folks who are not being responsible, who are not using as part of a workflow, but who are basically beginning and ending their creative process with the typing of the prompt and hitting the enter button on their computer or their phone or wherever they happen to be doing it.

While it's fair to say, in the best case scenario, yeah, people are using it in this way that is additive, I think it's fair for folks to be worried about the vast majority of folks who aren't going to take all those safeguards. I think there are some things that should hopefully give us pause or hopefully make us rethink if we're going to add this into our workflow, and if we are going to add it into our workflow, maybe to do it in the most responsible way possible.

AI tools are artificial but they're not intelligent

The biggest thing, I think, is that creativity. These large language models, we call them AI as kind of a shorthand, but they're not really intelligent. They're not intelligent at all. From my understanding, and please feel free to correct me in the comments if I'm wrong, what these large language models do is essentially just randomize a bunch of stuff. But what they're doing is they're randomizing all the stuff, but putting it through a whole bunch of pattern recognitions. And this is what all the data they've been ingesting from the Internet and books and everything else that they've thrown into these models. So they're able to take all this random stuff and sort of put it in an order that starts to look more or less like understandable English, because they have so much or whatever language you're using, they have so much of this data that they can, even though they don't really know anything, they don't know what the words mean, they don't know what their relationships mean, they can look at it and say, well, I've seen this a million times, and this is kind of how it tends to go as they're able to sort of put things in the right order. And then you kind of get stuff that seems to fit.

AI will lie or make things up

When it doesn't fit, they call those hallucinations. And I don't know if that's quite the right word, but that's what they call it. And this is when the AI will put some stuff together. And if you squint at it, it looks OK. But when you dig into the details, it doesn't look right. Something is wrong. Even if the pieces might kind of fit together, if you're able to analyze it, it doesn't work. Depending on what field you're in and how much how much experience you've had with using these large language models, you may have experienced this using it personally or professionally. I actually had my first big hallucination with perplexity AI, which is an AI that I use for search, essentially. And the only reason why I found the hallucination is because I'm not using perplexity to generate stuff. I'm generally using it as a search surrogate. So instead of going to Google, type in something, I'll go into perplexity and type something and then sift through that and then go through the sources. Well, it sent me on a wild goose chase. It had told me that there was a quote from a Roman author, Pliny the Elder in a book, and it had completely made the quote up. And I spent about three hours trying to track down this quote. Could not find it. I started off with the chapter in the book that it told me the quote was in. It wasn't there. I started looking at different translations. I started looking at different versions in case I thought that the Latin change or did the English translation change. Then I started looking through other books in the in the in the major work just to see if it did give me the wrong book. Then I started looking at, well, did it give me the wrong author? Bottom line was it was completely made up. And I was simply using the AI as a research assistant. I wasn't using it to generate any actual content. If you're not careful with the when you are actually asking to generate content and maybe even if you are careful, you might get something for a game system, let's say, that's wrong in subtle ways that you don't notice at first because it looks pretty good at a distance. It's only when you dig in. They don't may only be when you suddenly deploy this thing at the table that you realize something has gone horribly wrong or maybe it's a little something. Maybe it's a big something. We don't know. That's the hallucinations. And those are problematic. Now we're playing games. So you might think that it's no big deal. And it really in the scheme of the great scheme of things, it's certainly not a big deal. If you read the news and follow this stuff, you will notice that there have been times when the A.I. have made goofs like this in much more important fields.

AI is not accountable

You can't take the content for granted. When the content is wrong, where is the responsibility for that content? There's no sourcing. Sourcing is a big thing, especially in kind of the old school gaming, which I'm usually a part of. But even if you're playing other games, oftentimes, if I'm doing a video and I'm looking at something and I'm using a tool, someone will ask me in the chat or in a comment or in a message somewhere else, hey, Todd, I really like that. What what tables did you use? What tool did you use? And I can point them to it because these things were done by people. So I could say, hey, I was using the D30, you know, Dungeon Masters companion, or I was using Worlds Without Number, or I was using the A, D and D appendix. And they can go find those tools and use them themselves and repeat my results if they want to. You can't do any of that with the large language models. Maybe if you have a local model and you use your prompt exactly the same way, maybe you'll get the same result. But certainly with the stuff that people are using where they're online, the models are constantly changing. Even if you're using the same model and you're never going to be able to get the same thing. Try it. Tell me I'm wrong. Go in, open up a couple of these a couple of different windows with the same kind of chat, GPT client, but just maybe in a, you know, different isolated sessions, put in the same prompts, see what you get, see what you get back. And then if you get back something, so this is great. I'd like more than this. How can you guarantee to get more than that? More like that. If you're doing Worlds Without Number and you like this aspect of Worlds Without Number, you can go to that same book and say, great, I really like this aspect. Well, I'll use this aspect or maybe I'll find another book. Maybe I'm doing cities now. Maybe I'll do cities without number because, oh, look, they've got that book and this stuff will kind of fit together. Great. Because it's by the same author. There is a sense of responsibility who owns this, who did this thing. And I can share that source or those sources with other folks if I think it's a great source or if they think it'll be useful. And we can kind of all work using the same tools.

AI can't be sourced.

You can't do that with the large language models. I can say go to chat GPT. OK. Oh, well, here, let me give you my prompt. OK, but now I'm not doing it at the same time. It's it might be just different just randomly because that's all it's doing. It's putting random stuff together. This is also why you can end up getting similar, maybe different results. Keep asking the AI for names and you might get the same names popping up. Maybe sometimes it's a female name, sometimes the male name, maybe sometimes it's a dwarf name, sometimes it's a halfling name. But you might find that why? Because it's pulling from the same pool and ultimately it doesn't know or care that it's been giving you the same result over and over again. Even sometimes if you tell it something like give me something different, you'll still end up getting the same. And what is that same? It's just munged up stuff from this common pool that you can't then source. If I really like a set of names and like to know, hey, that name generator. Where did that come from? Because I'd like to use that. I really like what they did. I can't find it. It's a black box. This is a go to chat GPT, I guess, and hope, hope for the best. Good luck to you. Fair LLM prompt writer. Godspeed.

So you can't source it and nobody ultimately is responsible. You can't control the process. You can't share your process. You can't source the material. The way it gives you finalized results even takes away that interpretive power. I think there is a creativity in taking results that you're getting from tables or inspirational prompts and merging that into something that's yours. Somebody once said to me, I think it's a comment to another video or might have been a post or something I'd made that, hey, you know, these AIs, it's basically just like using random tables. And I disagree. I know there's some random tables that give you an entire encounter. Even then, I think Shadow Dark will do this, but even then I can take that and at least I can say again, following the can I source it? Can I share it? Yes. Does it have it? Can I see who's responsible for this for go to for ill? Great. It follows all those. But most tables are pieces and then you take those pieces and put them together like a collage, like an interpretation. And I think that's also where your creativity comes in. But large language models, as far as I've seen, they're giving you the end result.

They're just giving you the oh, well, here it is. There's a polar bear and it's on a nice flow and that gives you a stat block and it gives you what it's doing and all that stuff. And some people are looking like, oh, great, it saved me all this work. Yes, it has saved you all's work. It's also taken all the creativity away. And if you want to tell me that when you get those results, no, you tear them apart and you use your creativity to do this and the other then great. That is what I would hope you would do. I just don't have the confidence in people as a whole that that's what's happening. And from what we see of a product sneaking into drive through and other places like that and other formats or blog posts, I think we see already that a lot of folks aren't doing that work. So if you're doing that work and you're taking something and it gets you and you're spinning it and putting it to your own, that's great. I still think that the it's missing. It's going to miss that spark of your creativity. I think you're going to end up with stuff that all kind of looks the same because it's all coming from the same material. Maybe it's put into a blender and spun up and then spat out and put back together. But you're still going to see common elements coming up because all these AIs are drawing from common elements are the ways around that. Yes. Are most people going to be able to do that? No.

What do we do about it?

What's the answer? What do we do? What do we do with this? We can't put the genie back in the bottle. People aren't suddenly going to forget that these things exist. They're not all suddenly going to be pulled down from the marketplaces. What we can try to do is try to be responsible. What we can do is understand when people like me and many others are very hesitant about that or are not. Inclined to be enthusiastic about the adoption. That's not about being, oh, you're just a load. I hear technology. No, I think it's you have tools and you have to try to use the tools for sure what they're best made to be. If these the tools to be the to do the thing that the tools made to do. But there could be a point where the tool stops being a tool. If the tools doing all the work, is it a tool anymore?

Go to the movie 2001 and when I guess spoilers, but it's from 1960, what, 869. So sorry if you're super late to the party. In 2001, the second part of the movie or whatever part where they're heading off towards Jupiter, I guess, or wherever that they've they discovered the monolith and you have how and you have the humans are on board. How was designed to be a tool? But how comes aware, self aware and starts to try to kill off the people. They don't really discuss why, at least my recollection is then it's not discussed in the movie. Why how does this? Maybe it is in the book. I don't know. I haven't read the story.

But it's clear that the humans have become really just an adjunct to how it's almost like the humans are now the tool to how doing everything. The the whole ship and everything is completely under the control of how. And the humans are only needed to sort of support how it's kind of flipped. And I think the movie is trying to tell us something about tools and nature of tools. And when tools become too much and they try to destroy us, we then have to figure out to overcome our own tools, because the tool is no longer the tool and almost it's almost the roles have been reversed. We go from using the tools to the thing using us as tools.

On the one hand, you look and say, well, you know, technology and we understand more this than the other. The other thing they're the part of is, hey, we we want to keep this a certain kind of game. We want to keep the game in the certain parameters. Otherwise, at some point, not the game anymore. It's something else. And I think it's fair to ask in tabletop role playing games. Hey, one of the things that has defined tabletop role playing games is the human aspect. What separates tabletop role playing games from computer role playing games?

These tools at some point start to draw that humanity away. We go from the humanity being the center to being the manager. And maybe at some point, the humanity just becomes the tool for the LLM or whatever computer thing we're doing. I know I've seen things people have talked about. I've read about it on Reddit and probably elsewhere where they've had dungeon masters who are basically taking all the things that a party is doing and inputting it into a large language model to then spit out how the world responds. At some point, maybe it's already there. You could just put a mic, turn the mic on your computer, have it just do the voice to text. And do you need the dungeon master anymore? And some folks will be thinking, oh, that's awesome. So we can all play it. Nobody needs to be the GM. OK, but then it's not a tabletop role playing game anymore. To me, it's a computer role playing game. You're just not playing on a screen. If I want to play a computer role playing game, I'll play a computer role playing game. But if I come to your table, virtual table, in person table, ultimately I want I want your creativity. I want your game. And for me, the more we rely on these large language models and other A.I. to determine our game game for us, the less the games are really ours and the games are theirs. And we've gone from being the creator to the manager. And I don't need a game managed. I want a game created. So that's where I'm at.

Let me know what you think. I'm sure folks will. It's if people have very strong opinions about this. I am curious if you use these tools in your games. Let me know how you use them and let me know if you care about how they fit into your workflow.

I know they can be time savers. I know that they can do a lot. That doesn't mean we shouldn't ask questions about the whys and the wherefores and where it all leads at the end of the day. I know we're all in in this to have fun. And I don't mean to say your fun is bad wrong, but there can be implications or maybe we should just talk about it differently. Maybe we should just call it another game so that we can have some of us who are playing tabletop role playing games that are human and others are playing computer tabletop games or, I don't know, assisted. I don't know. And yes, there are some gray areas or probably some fuzzy zones we can get into. But hopefully my point stands or at least it's something to think about and I welcome the discussion.

With all that said, game on everybody and I will talk to you later. Bye now.