Transcript
Speaker: In three years, Cyberdyne will become the largest supplier of military computer systems. All stealth bombers are upgraded with Cyberdyne computers becoming fully unmanned. Afterwards, they fly with a perfect operational record.
Speaker: The Skynet funding bill is passed. The system goes online on August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate.
Speaker: It becomes self-aware at 2.14am Eastern Time, August 29th.
Speaker: Come get your tune-up It's time for your tune-up Your game needs a tune-up Come get your tune-up It's time for your tune-up Your game needs a tune-up
Speaker: Welcome back, everybody, to the Tabletop Tune-Up. I'm Mark. I'm here with my friend, Ben. How you doing, Ben? I'm doing well, Mark. Looking forward to this episode. We are doing something that I don't think we've done in the past. We are revisiting one of our topics.
Speaker: Oh, that's right. We are. and Was that intro a little too dark for you? I don't know. well Feels interestingly retro, not just because the 1997 reference, but because like, yeah, i kind of think that's happening right now.
Speaker: You know, folks, what a drone warfare out there. Folks, we're coming back to talk about AI. It's been a year since we first brought this up. And I want to point out, it was our first episode that got a thumbs down. And I'm hoping we get a few more this time.
Speaker: We want more of those thumbs downs. That's right. So in this episode, we're going to revisit some of the concerns people have. We're going to talk about what's changed a little bit and maybe more importantly, definitely more importantly, we're going to talk about how we're using AI differently than we were last time we talked to you about this. Yeah. A year, it seems like honestly a lifetime ago in the world of AI, doesn't it?
Speaker: Things have happened. Yeah. I work with a lot of this stuff at work. We're analyzing and trying to figure out like, what can this be used for? Can it be made valuable to us? And the answers are mixed. It's not like there's just one clear answer. It does all things better and great, wind it up and let it go. and And who doesn't love training the person who's going to take over your job?
Speaker: Everybody doesn't love that. Yeah. So we're going to start actually, Ben, let's let's talk about some of the concerns with AI real quick. I know we touched on these the last time we did it, but I don't think those concerns are going away anytime soon.
Speaker: um So these are like some of the concerns we hear constantly from from folks when we talk about Bye. Yeah, and i'll I'll just say this is also stuff I'm seeing a lot in the working world too. ah Companies that are adopting AI are having these concerns. So it's interesting that it kind of crosses that boundary, that everybody's concerned about it, whether they're using this for personal use or at work, whatever it might be. The number one thing there is data privacy.
Speaker: So let's imagine that I'm uploading some campaign notes into one of these models, a chat GPT or what have you. People then wonder, am I losing control of my intellectual property here? Do i lose ownership of my campaign ideas if I upload them?
Speaker: Valid question. Ben, do you know the answer to that? Let's say you upload your campaign synopsis, your realm and all that stuff in the chat GPT. Does that now owned by Google?
Speaker: So I think the answer, generally speaking, is it depends on the model. And in general, the answer is no. um Interestingly, though, I've seen some things that say if you are using AI to produce things, ah then those things that the AI produces may not be patentable.
Speaker: So it might turn out that if you use AI to write up your campaign world, that you might have some problems if you thought, hey, I'm going to turn that into a published product. Really?
Speaker: That's wild. i don't I don't know if I like that. No. well definite i think I think part of the thing, though, right, is like you need to be able to show your authorship and your contribution to it. So I think the interesting gray area is mixed stuff. And I certainly think that there's a lot of people working on this.
Speaker: This is kind of a question that's a lot more relevant to... I think image-based AI. We had Terminator in the in the opening sequence there, and James Cameron's actually in a lawsuit right now, and people are challenging whether or not he can legally use image-based AI or whatnot.
Speaker: Cameron's interesting defense is that he says... look, don't judge on the inputs. Don't judge on how the model was trained. Judge on the outputs. If you think it's plagiarism, then let's judge what the model produces. And if that is plagiarism because it has too great a resemblance to its sources, then let's treat that like everything else.
Speaker: Boy, it really sounds like the future... If you're a young person and you're going to the future, copyright law sounds like a wonderful director.
Speaker: Copyright law and AI because that's going to be – this stuff is all going to be litigated. Yeah. I think we're very much in the early days of the internet with respect to this kind of thing. If you remember of the nineties and Mark, I know you do. ah They had a lot of these same kind of questions and concerns about what happens when you put content online. What happens when I put this or that image online? What happens when I put this or that text online? Does that mean that everybody, everybody can in principle at that point, just yeah grab it.
Speaker: ah And you remember, there's a lot of concern over what happened in intellectual property then. But they eventually kind of worked that out. And I think they'll work that out here too. Well, also, there's a ah term I've heard before, and I think you brought it up as well. Cognitive offloading.
Speaker: Yeah. Where the stuff is dumbing us down, um making us stupid. Right. Over the last summer, there was a study published by MIT. And what they found was that students who were using AI to write papers had a harder time quoting from those papers and didn't really know what was in them with the same degree of conceptual clarity as people who had not used AI or who had used it in a slightly different way to kind of brainstorm, but they actually wrote the text and went through it all.
Speaker: ah And so cognitive offloading is just, this is actually a high a highfalutin term for a simple idea. um Mark, I'm, as you know, not good at math. I use a calculator to cognitively offload the arithmetic that I would do in my head to a calculator. And it makes sure the sign changes right and it can do the big numbers and so on.
Speaker: Fair use. Cognitive offloading is something we also do when we use a pen and paper to write notes. We're offloading our memory need. So all of these things are just different things that let you use technology to not do something in your head. And the concern is, well, look, if I'm using AI to cognitively offload creative tasks or analytic tasks, is that making me dumber or less creative?
Speaker: Yes.
Speaker: Honestly, I'm cognitively offloading my entire life. So, you know, Yeah. Well, you've been using ii for a while. So, i mean, what's your perception? Do you feel like less creative or do you feel as though AI is becoming too much of a crutch in your process?
Speaker: Well, I think we can get it into that a little bit later, but the answer is no. have But I want to jump into that a little bit later, if you don't mind. Okay. Let's talk about, since it's been a while, it's been a year, what have been some of things that have changed in AI? Yeah.
Speaker: Well, a couple of things are different. One is that we've got new models coming out right now. As of this ah podcast, chat GPT 5.1 has come out. And in general, one of the things that's really interesting about the five series models from OpenAI is that they're quote unquote thinking models. So they will take extra time to review what you've asked them. They might even ask you follow questions about it. And they're taking extra time to kind of refine their outputs and make sure they're answering the question that you asked. So not really a anything specific to our games and stuff here, but the point is it will probably give you better results if you haven't tried it lately. um Maybe try it out and see if it's better now.
Speaker: Contrary, it is very much something that's being useful in the games. When I've written up a thing for a game, oftentimes it'll give me a prompt afterwards.
Speaker: you know Would you like to use this as a campaign material? Would you like this rewritten as ah material to print to your players? like It'll literally ask me ah something related to a game, which is remarkable and yeah amazing.
Speaker: Yeah. um So that's one of the things that these new models will do is they they know they've got to keep the conversation going. And so in the background, they're always going to find some new question or ask you if you want them to continue in some way. So that's really interesting. and I think, you know, if you're somebody who feels you know conversational with these things, then you'll find a lot to do with them. Another thing that's new and interesting is that the image models out there are more sophisticated and people use them to do a lot of photo editing. They use them to redraw or stylize things in ways that we used to as well.
Speaker: Here it's interesting, I think, because a lot of times we're used to seeing the AI image models. You can kind of tell when you're looking at an AI image these days, for the most part. Some of them are very realistic, but but we're seeing a lot of AI imagery out there and on the internet, and we're learning how to detect it.
Speaker: I think what's interesting is that one of the things we haven't seen people talk as much about is how AI can do editing in Photoshop, for example, in ways that used to be very painstaking and hand done. And I don't see anybody complaining about that. Like the people that I know that are working in graphic design, you know being able to kind of instantly correct this or that thing or extend that picture into a 16 by nine aspect ratio from a four by three, they're pretty happy with those kinds results.
Speaker: Yeah, I did something kind of interesting with the image model with ChatGPT recently. I had a battle map. um It was a big battle map. And I wanted player version of that battle map or that top-down topograph map of this area.
Speaker: And I just followed the image and say, make a hand-drawn, crude version of this. boom. And it made it. And it it was, it looked like, it looked like a child drew it, which is kind of what i was hoping.
Speaker: um And it saved me hours of futzing around trying to make one myself. Yeah. I think what's interesting about that example is you're using an existing visual asset that you've created and you're asking it to edit and restylize that thing into a different kind of visual presentation. So interestingly, it's not like you're asking it to start from scratch and build an image, but you're asking it to edit for you.
Speaker: Did you try multiple versions of that when you when you did that? did you Did you say, try it once a sepia thing? and It spit it out the first time and I was fine with it. um And I just moved on. I mean, what I kind of like to use this AI for is to save myself time at at the table.
Speaker: Saves you some prep time. What ah what is this about a multimodal AI? what What the heck is that? Yeah. So multimodal AI ah originally came out or was supposed to be a part of chat GPT-4, but the core idea here is that it's AI that can operate simultaneously in multiple modes. So for example, images as well as text at work, for example, we gave it like a flow chart that had all these boxes and arrows and said, could you please rewrite this as a, like a, an actual step-by-step process. And it could do that.
Speaker: And it did a pretty good job. um In our case, I think what's cool about multimodal AI is you could have it, for example, write box text when you feed it a battle map of an encounter area, for example, or you could have it generate some combat mechanics for that battle map area based on what it sees in the environment, right? So if there's a difficult terrain thing over there, if there's a mechanism that maybe ah you know could be a trap or something like that or a puzzle. it could It could write that stuff up in language, having seen it as an image.
Speaker: And i think that's a really, really fascinating way to combine those capabilities. So that's a new thing AI can do now. Amazing. Every time i I touch this stuff, I'm like, what is this magic?
Speaker: It's so And I think you started getting a little more exploratory in your use of this. You're using it a little more now than you used to. So yeah tell me about that. So last time we met, last year, i had just been writing up something and throwing it in a chat GPT and asking it to rewrite and clean it up.
Speaker: So it would spit it back to me and I would use that, like I would do most of the work there. So now what I've been doing is this last game I've been running, I've uploaded my entire campaign information.
Speaker: I think about eight pages of campaign material. This is just a write-up of the actual campaign, realm information and lore. And then I uploaded a character description and for every single character that's going to be in this, the main characters.
Speaker: And that was just a start. As the game goes on, I just keep asking it questions, asking it for things. Can you write a description, a box text of this ziggurat? And it knows the ziggurat because it's got my campaign material. And it's able to explore those those places and write about them with knowledge that I have provided to it.
Speaker: Yeah, the ability to have kind of instant information or instant generation is really interesting because... That's something that individual GMs might be better or worse at. And so having that there is great. ah One of the things that we're seeing people can do now is they can create specialized agents that are always trained to do specific things. So you could have an AI agent that is has got instructions already kind of pre-configured so that it knows...
Speaker: um how to write box text or how to write a monster stat block and how to write NPC dialogue, for example, if you fed it your NPC information. yeah So those are fascinating kinds of things where you can give the AI persistent training and data context is what they call it. So that it always knows how to do certain kinds of tasks. And it can be really helpful for you both in the moment and in prep.
Speaker: I've even once uploaded a description of, in Forge of Fury, the Glitterhame. And this is no joke. I uploaded the description of the Glitterhame because i was getting ready to for my game the next day, and the GM that was going to be playing in that game posted the image of one of the creatures he was running against the the group of players that that are in my game.
Speaker: And it was the same creature i was going to have them fight in that Glitterhame. And i was like, oh no, i have to come up with the new creature for this environment. And I have like, you know, 24 hours, you know, and I'm like, dang it.
Speaker: So I just uploaded the the description of the Glitterhame into chat GPT. Right. And this is this is an underground cavern, right? Yes. This is in Forge of Fury. It's this real this cavern with all these gems and this fungus at the bottom. It's glowing. It's very alive with light. And I said, make a creature...
Speaker: based on the environment that they could fight in this environment. I just wanted to see what it would do. And it made this, this beam of light that would come out of the lights from the glitter hame and coagulate into a solid light of some sort.
Speaker: And it had like its special attacks, it had its, you know, its actions and they had it all written out. And i was like, What the hell? Instant monster generation.
Speaker: And it was, I told it what level it was to be. And it was appropriate. Like it knew kind of the level for the encounter, how many people were in the encounter. And I was, I just made a creature.
Speaker: Fantastic. yeah gp that I didn't do anything. Well, no I mean, this is interesting, right? The role that you're in is, is not a creative role. It's an editorial role. So you said to ChatGPT, this is what I want you to produce. And it gave you that.
Speaker: You could have, at that point, whenever it gave you the first draft, you could have said, i like this and I like this, but change that and change that. And it would have rewritten it and incorporated those changes. And so it's just like being like a newspaper editor and some journalist comes to you with the story and you, you know, shape that a little bit. I don't think it's something I'm going to use a lot, but um I'll tell you, it was useful in the crutch. You know, like I was stressed for time and it it it did the job for me.
Speaker: So you've got a really interesting program you've been kind of messing around with a little bit. Do you mind talking about Session Keeper AI? Yeah. Yeah, so this is ah this is beta software. So I'm not going to necessarily ah talk a whole lot about it. But I did find it really interesting. There's a developer who's created AI that will take a session recording. So imagine just turning on a recorder at your table. And it will create a transcript of that. and it will identify voices with characters, and then it will create a session overview and provide all kinds of really interesting details and notes and analysis of what the players are doing. And it will convert that into insights and open questions and session recaps and a whole bunch of really interesting stuff that
Speaker: Frankly, if you're in the session and you're a good note taker, probably you're fine. You don't need that. But I am not a good note taker. So I really appreciated having that to ah to look back on, to call back to. i liked that it was giving me player specific, character specific, I should say, specific insights about where things could go, what kind of monsters or encounters I could try, what players were doing well and so on.
Speaker: And for me, like I'm hearing you talk about this. I'm one of those GMs who makes so much stuff up on the fly. This includes names and places. I'm just creating them up out of the out of my head as we're gaming.
Speaker: What an amazing tool to have that would write that down for me. um If I don't have a player at the table willing to make those notes, yeah this could do it. That's really cool.
Speaker: And it also then generates a basically a wiki for that campaign too. So every person, location, event, all of those things get documented into the session wiki. There's a lot of really fascinating stuff that it's producing. And ah I think this is not the only note-taking and transcription software out there, but sure this is this is a fascinating kind of thing. And I think if you haven't tried out stuff like this, friends, if you're open to it, check it out and see what you think. It can save you a lot of time. I think that's kind of the the headline for all this stuff is that all of this is designed to save us time. And if you think about what that time savings can get used for, then it's playing with your friends.
Speaker: And also guys, don't, don't let yourself get lost in the AI. Like again, I mentioned this the first time we did this, make sure your voice is heard, make sure you're writing as much as possible that to give it your flavor, your flair And because that's what people tune in for. That's what people want.
Speaker: They want you. They don't necessarily want the AI. However, these tools can really help bring your voice out. If you're a person who's not a very good writer, it could really give your voice some extra weight and carry it further. So just be thoughtful, thoughtful when you're using this stuff.
Speaker: Yeah. Treat it like a partner, not a expert. Yeah. Yeah. Ben, what are some of the predictions you might be thinking of for the future of ai at the game table? Yeah, there's ah there's a guy who studies this stuff over at Wharton. And one of the things he said, really fascinating guy, if you're interested in AI, start reading Ethan Mollick.
Speaker: But he said, keep in mind that any AI you're using now is like the worst AI you'll ever use. And he didn't mean that it was bad. What he meant is if I said that to you about your personal computer back in 1980, whatever, 90, whatever, whenever your first computer was. Fight words.
Speaker: Yeah. Well, but every computer you've had since then has been better. And in the same sense, every AI you're going to use in all the years to come, it's going to have some improvement. It's going to have some new thing about it. And so I think what's interesting is we're going to see AI that becomes more specialized. It becomes more useful to us. And I think the real trick right now that everybody's trying to figure out is how do I not do the wrong kind of cognitive offloading, right? That thing we talked about earlier. How do I not get stupider and less creative because I'm giving it things that I should be keeping for myself?
Speaker: Yeah, that's a good that's a good point. How about you, Mark? Yeah, what do what do you think the future holds? I think I'm going to just go with what our friend Randy had mentioned earlier. i was talking with him about AI, is that we're going to have ah this stuff like Session Keeper. It's going to record our sessions.
Speaker: It's going to create an audio for it or an animation. and we're going to be able to watch our sessions in an animated feature. Oh, that'd be fun.
Speaker: After a session. I don't think we're that far off from it, to be honest with you. think you're right. Well, with that,
Speaker: and it is time for this week's tune-up segment. Mark, this time I've got one for you. What'd you got? This week, I got disinvited from a game. This actually happened, for real. I got disinvited from a game.
Speaker: They were playing a game that... ah I will not name. It was ah kind of a more technical, crunchy game. And I think they had their kind of vibe about how they wanted to play it. They were, they kind of met offline. I'd played two sessions with them. um They all kind of agreed the GM and the other players. And then the GM kind of broke the news.
Speaker: Very gracious, totally fine. I still feel a little bad. Like, you know, maybe there was something I could have done differently, but question for you is for anybody that's in this kind of situation, do you try with the group again? do you try to suss out what you did wrong? Like,
Speaker: What do you think the future is for somebody who's had that experience? First of all, Ben, you're going to tell me who this person is. going to go talk to him right now. That's a friend indeed. Yeah. Having been on the outs before too, and having to put people in the outs.
Speaker: Sometimes you just have too many players at the table. we've We've had to tell people that we don't have room for everybody and it's it's never fun. So like, i I feel your pain. Don't try to take it personal. Yeah.
Speaker: It sounds like they were being gracious. They probably had good reasons. It could be scheduling conflicts. could be there's too many people at the table. Who knows? you know say Maybe they express that to you. But there's worse ways to be kicked out of a group.
Speaker: Yeah. And you've got a nice, gracious way. And in that note, I would say i wouldn't give up on that group. I mean, maybe you give up on that game, but there will be other groups, other games that come along. And I would say apply for that game if they want. Yeah. Yeah, I think that's kind of the the perspective I'm taking on it. um i think they were all good dudes and I would play with them again, but I don't think I'd play that game with them again because I think yeah it just was not a good fit for the kind of gameplay I'm i'm good at. And friends, if you're out there and you've had experiences like this, it feels a little bad, but I think that's about it.
Speaker: We've had this kind of situation in our our table where we've had, you know, we got a ah pretty good game, pretty big game group on our online discord. And every time we start up a game, we know we can't take everybody.
Speaker: Yeah. And that's a challenge every time. So folks, it doesn't mean you're not friends. Just um there's always other games. Maybe start up another game another day of the week and you'll get a few of those players come join you. So there's all different ways to handle it.
Speaker: I think with that, we're going to wrap our episode here. See you in a couple more weeks when we'll talk more about your ideas and your thoughts. If you have a tune-up segment you'd like us to think about, or if you'd like us to ah ideate on some new ideas you're having, go ahead and send them to us at tabletoptuneup at gmail.com or let us know in the comments if you're doing this on YouTube or one of the other platforms that we do this on. Love to hear from you.
Speaker: Until next time, keep those dice rolling.
Speaker: We'll show you
Speaker: Level up your fun to know
Speaker: Your quest has now been
Speaker: True love.
Speaker: Come get your tune-up It's time for your tune-up Your game needs a tune-up Come get your tune-up
