It's debatable whether everyone wants the tech to get better.
You're going to find it's a hot debate about whether the social elements of DnD should be limited or reduced, or whether this sort of enterprise is best left to explicit video games.
I dont hate the idea of AI powered GM aids, but even that gets into murky, ethically dubious territory when you start looking at what you'd have to train them on.
1) I think framing the issue as a zero-sum game between "increasing the capabilities of AI" and "reducing the social elements of DnD" is a false dichotomy to begin with. The former does not require the latter.
2) While training AI can involve a lot of ethical quandaries, in this case there are none - it would have to be trained entirely on games run through the DDB service or any partnering VTTs, which all of us would be explicitly agreeing to by using those. It's not like they'd be able to monitor meatspace games even if they wanted to.
I feel like this would end up being the worst of both worlds between TTRPGs and video games.
You have the limited options of a video game, where the devs can only account for so many factors from the players. Even games where you have choices on things, tend to limit you to a few. On the other hand, you don't have the full on production value of a game like Baldur's gate or Divinity 2 Original Sin.
I wonder how it would even train full sessions. A lot of people playing online use discord or other programs to chat for example, how do you process that? Or do you try and feed it without the context from voice chat? How do you get it to incorporate homebrew or house rules?
I could see an AI handling the pure war game, RAW aspects of the game, but struggling with even things like adapting the AI of how say, a swarm of goblins approach fighting the party as opposed to a battalion of knights. Let alone aberrations and dragons and fae etc. Just making the fights feel dynamic and not like you're plying out a random encounter in a turn based JRPG on a grid feels like it may be difficult for an AI to authentically reproduce.
If this actually happens, then good luck to them. I'm skeptical but I'm not necessarily against the idea as long as people have to opt in to having their sessions fed into the AI.
I could see myself using an AI DM for my SO and I to play by ourselves between sessions of regular campaigns. I’ve tried games like Gloomhaven, but, between the scant actual roleplaying and over focus on mediocre and repetitive combat, I just could not get into that style of game. A competent AI DM that can truly engage in all aspects of D&D is certainly something I might give a shot.
Now, I also do not necessarily have confidence that an AI DM will ever truly be “competent”—there’s a lot that goes into good DMing that comes from reading facial expressions, tone of voice, conflict mitigation, external conversations with parties, being willing to abandon entire plans if they are not working, etc. that AI is not presently equipped to handle. I am not sure some of those problems can ever truly be handled by a computer - but I would be curious enough to at least give them a shot.
I just saw the slightly negative comment on Gloomhaven and got a bit mad. But otherwise, I agree with what you said. Gloomhaven pretty much satisfies my 'need' for the combat bit of DnD and other TTRPGs, so I mostly play DnD to roleplay. I have a sneaking suspicion that A.I dms will never really be able to be as good as an actual DM, but who knows.
I could see myself using an AI DM for my SO and I to play by ourselves between sessions of regular campaigns. I’ve tried games like Gloomhaven, but, between the scant actual roleplaying and over focus on mediocre and repetitive combat, I just could not get into that style of game. A competent AI DM that can truly engage in all aspects of D&D is certainly something I might give a shot.
Now, I also do not necessarily have confidence that an AI DM will ever truly be “competent”—there’s a lot that goes into good DMing that comes from reading facial expressions, tone of voice, conflict mitigation, external conversations with parties, being willing to abandon entire plans if they are not working, etc. that AI is not presently equipped to handle. I am not sure some of those problems can ever truly be handled by a computer - but I would be curious enough to at least give them a shot.
I just saw the slightly negative comment on Gloomhaven and got a bit mad. But otherwise, I agree with what you said. Gloomhaven pretty much satisfies my 'need' for the combat bit of DnD and other TTRPGs, so I mostly play DnD to roleplay. I have a sneaking suspicion that A.I dms will never really be able to be as good as an actual DM, but who knows.
If this actually happens, then good luck to them. I'm skeptical but I'm not necessarily against the idea as long as people have to opt in to having their sessions fed into the AI.
Welllll.... Not to be that guy, but you already agreed to give them access to any of your usage data for any Website, Game, or Service they offer, just by posting here or using any of their online sites/tools. We all did.
If this actually happens, then good luck to them. I'm skeptical but I'm not necessarily against the idea as long as people have to opt in to having their sessions fed into the AI.
It's an opt-out system --
we opt to play D&D on a VTT that is out in the community. ;)
Rollback Post to RevisionRollBack
J Great Wyrm Moonstone Dungeon Master
The time of the ORC has come. No OGL without irrevocability; no OGL with 'authorized version' language. #openDND
Practice, practice, practice • Respect the rules; don't memorize them • Be merciless, not cruel • Don't let the dice run the game for you
WOTC IA would not be able to run as a DM in a game session under the current rules. I have a good understanding of this from what I used to do. BG3 in it's turn assisted mode needs a really large install on a very expensive computer to do what it does and it is limited to a very restricted rule set, not the current rules.
WOTC and HASBRO are aiming at a presentation layer running in a web browser on a computer, tablet or cellphone like the current D&D Beyond character sheet.
That is never going to happen within the next 5 to 10 years on a computer or really powerful laptop in a web browser and certainly never on a cell phone in my lifetime even if some clueless executive at HASBRO thinks otherwise with the current rules set.
The 6th Edition of D&D would have to be gutted massively from the current 5th edition to get it to work. Think the difference between pong and BG3 where 6th Edition would be pong...
A.I. DMs aren't a terrible idea, however I'm pretty damn sure that what ever A.I. WotC try's to give us wouldn't even be half as good as AI dungeon which itself couldn't replace a real quality DM. The reason why I don't think its not a terrible idea is because there is a DM shortage in 5e. though I have heard there's a player shortage in other rule sets.
I would agree to some small degree. I think that an AI DM could have some use. It might be good for a very niche use, even if they don't generally replace human DMs. In a situation where there is no way in which you could have a human DM or you want a solo experience, for example. However, there are two main reasons I think they will never actually replace hman DMs:
1. A large part of D&D for many people is the human interaction and the shared experience created. It's not just about telling a story, after all I could write a book if that's what I wanted. Nor about the mechanical gameplay, I could play a regular board game if that's all I wanted. Yes the fusion of these two is part of what makes D&D so awesom, but so is the fact that it creates a shared experience between real human people.
2. This is an even bigger issue: DMs can give very tailored experiences. There's so many ways in which a game can be subtly manipulated so give thousands of different possible experiences. An AI can have "settings" like "comedic," "dark," "heavy on RP," but will never be able to do what people can, at least with current technologies. AIs take absolutely gargantuan amounts of data to learn compared to humans, so it simply isn't possible for them to learn about the preferred play style of the player(s) in a feasible amount of time.
You know what they call an A.I. trrying to run a game?
A video game.
If I want to play video games I'll play video games. I have zero interest in an A.I. clumsily failing to run even the simplest of D&D games.
As someone who works on AI, this is simply not true. There is one critical difference, and one that makes massive differences. An ordinary video game uses conventional algorithms. That is, the programmer creates basically a set of "rules," which the program follows. If you do X, the game responds with Y. However, in a machine learning system, the programmer creates a structure to learn from data (usually a neural net but their are other approaches like rule based learning to decision trees), which allows it to find patters in the data. In can then, knowing this pattern, find the same pattern in thing it hasn't seen before. Oversimplification, but that's basically how it works.
So in a video game, everything that can be done has to be programmed by the creators. The player can only affect the plot to a certain degree because the developers have to have created all the possible endings and so on. The players can also only take certain actions the the developers integrated into their code. The list goes on.
Using a machine learning system on the other hand allows the program to react in new ways that were never envisioned by the developers. If you do something new, it can significantly change the plot based on your actions. RP and dialouge can be real, as demonstrated by chatGPT and others before it. You can find new solutions to problems and make your own decisions. With a sufficiently advanced ML system, there may not even be a detectable difference between an AI and a human (which has already been shown in published research in some other contexts such as conversation and composition). The main difference is not the actual game itself but the human interaction that is part of it.
I could see myself using an AI DM for my SO and I to play by ourselves between sessions of regular campaigns. I’ve tried games like Gloomhaven, but, between the scant actual roleplaying and over focus on mediocre and repetitive combat, I just could not get into that style of game. A competent AI DM that can truly engage in all aspects of D&D is certainly something I might give a shot.
Now, I also do not necessarily have confidence that an AI DM will ever truly be “competent”—there’s a lot that goes into good DMing that comes from reading facial expressions, tone of voice, conflict mitigation, external conversations with parties, being willing to abandon entire plans if they are not working, etc. that AI is not presently equipped to handle. I am not sure some of those problems can ever truly be handled by a computer - but I would be curious enough to at least give them a shot.
I just saw the slightly negative comment on Gloomhaven and got a bit mad. But otherwise, I agree with what you said. Gloomhaven pretty much satisfies my 'need' for the combat bit of DnD and other TTRPGs, so I mostly play DnD to roleplay. I have a sneaking suspicion that A.I dms will never really be able to be as good as an actual DM, but who knows.
*The best AI DM online at the moment sucks ngl*
It's true, but then again it hasn't got even close to the amount of research as other ML applications, whether that be image recognition or even AI legal judges. I don't think that now is the time for an AI DM, but I don't think its impossible. I don't think it would be as good as real D&D, but it could be a lot better than AI dungeon.
Like monster choices. The Dm could input the players levels and classes and the Dm-Ast would list out all the possibilities. The DM would then just click on what they want and the stats would be posted and kept track of.
How about a list of normal NPC's to populate a town. Just input the towns size and shops can be listed and proprietors randomly named.
All of this could be based on settings and town sizes.
We do not need infinitely random mapping. A growing list of maps for each town size. The Dm fills in the details. The same for underground settings, Castles and other lairs. I have found that truly random map makers jest get it all wrong.
I'm with Yurei on that one. An AI DM becomes a video game.
The whole point of having a DM is to handle whatever the players may come up with where a computer game is built to manage and thus limit the outcome in its own boundaries so that it can react to what has been predicted by the programmers. Remove the players creativity and you've removed the purpose of the game itself...
If some of you have played Bloodbowl 1 and 2, you have probably realized how terrible the AI is and they have been working on it for years. I could even argue that the 95 dos game had actually a better AI. Compare this to what D&D would require, it is not even thinkable if you ask me.
If you'd want to build an AI, the first thing you'd want to provide is how to handle the Rules but also every possible scenario how the rule may be affected by other variable that may factor in. That alone would be a piece of work and it is not always obvious as there are many question that ends up being answered on twitter or else. That alone could provide a great tool for a GM.
The second thing that comes to mind building an AI, is if you're planning on working in specific settings such as The Forgotten Realm, being able to give the lore of every region including cities would be in itself a great piece of art. That again, could be a great tool for a GM. The party decides to go to Longsaddle suddenly and you're wondering what you're supposed to describe, type it in on the AI and you'd have something in hand to make it closer to the lore.
I'm with Yurei on that one. An AI DM becomes a video game.
The whole point of having a DM is to handle whatever the players may come up with where a computer game is built to manage and thus limit the outcome in its own boundaries so that it can react to what has been predicted by the programmers. Remove the players creativity and you've removed the purpose of the game itself...
If some of you have played Bloodbowl 1 and 2, you have probably realized how terrible the AI is and they have been working on it for years. I could even argue that the 95 dos game had actually a better AI. Compare this to what D&D would require, it is not even thinkable if you ask me.
If you'd want to build an AI, the first thing you'd want to provide is how to handle the Rules but also every possible scenario how the rule may be affected by other variable that may factor in. That alone would be a piece of work and it is not always obvious as there are many question that ends up being answered on twitter or else. That alone could provide a great tool for a GM.
The second thing that comes to mind building an AI, is if you're planning on working in specific settings such as The Forgotten Realm, being able to give the lore of every region including cities would be in itself a great piece of art. That again, could be a great tool for a GM. The party decides to go to Longsaddle suddenly and you're wondering what you're supposed to describe, type it in on the AI and you'd have something in hand to make it closer to the lore.
This is a fundamental misunderstanding of what AI is capable of. What makes AI both so useful and so revolutionary is that it can react to situations to programmers never thought of or hardwired into the system. It's called "machine learning" for a reason. The AI literally learns for itself how to do all this, usually by looking at data (though there is also reinforcement learning which still in many ways an emerging technology). For an AI to work, all you need to do is give it large amounts of data and give it the simple equations that govern back propagation (they're basically just the chain rule from calculus). If you don't believe me, check out chatGPT, word2vecc, midjourney, or even alphazero: they all can handle basically anything you throw at them, even weird and strange things that the developers would have never imagined yo using it for.
You're comparing the general concept "AI" to something that could in the loosest definition be considered an AI but if far weaker and more dependent on human inputs then the current AIs that exist. There aren't any really great AIs in the field of role playing mainly because it's not that important in context of all the other things AIs are used for. There just hasn't been a lot of serious research on this.
But even what has been made goes beyond this generalization of AI you are making. Comparing AIs to gloomhaven is an unfair comparison because gloomhaven is both a board game which cannot access the power computing power of a computer, and thus could never actually have an AI (ANNs take crazy amounts of computing power to run), AND never was advertised as being "an AI RPG." If we look at actual examples of AI RPGs, one of the best examples is AI dungeon. Sure, it's not great and not D&D, but it's definitely not a video game, at least not in the way as in it is not like D&D. It can react to most player decisions and change the plot accordingly. It doesn't just follow input programmed by the developer and can actually tell a cohesive story in which the player has the driving force and can significantly alter the plot in any number of ways, just like D&D. It's not meant to be D&D and has a ton of room for improvement, but I think it shows that an AI DM is possible. With a lot more research, effort, and funding I'm sure an AI DM that can act fairly similarly to a normal DM is possible.
I encourage people to look into those old epic text-based video games and tell me that a digital DM is different.
I will, but I can tell you before even starting that AI is a rapidly evolving field which has seen gargantuan innovation even in the last few years. AIs like StableDiffusion, ChatGPT, and AlphaZero would have been thought impossible in 2010 even. Plus with rapid advances just in the computing power, deep neural networks were basically just made possible in the 2000s. I think the term "AI" is often used for something that is not really what AI is capable of, leading people to misunderstand the difference between something that can learn on it's own like an ANN and something wherein the programmer must input all the rules for the AI to follow manually like for example most text-based games I've seen before.
What games do you recommend me looking into? I've never seen a text-based game that leveraged machine learning, but maybe there were some I never knew about.
I encourage people to look into those old epic text-based video games and tell me that a digital DM is different.
I will, but I can tell you before even starting that AI is a rapidly evolving field which has seen gargantuan innovation even in the last few years. AIs like StableDiffusion, ChatGPT, and AlphaZero would have been thought impossible in 2010 even. Plus with rapid advances just in the computing power, deep neural networks were basically just made possible in the 2000s. I think the term "AI" is often used for something that is not really what AI is capable of, leading people to misunderstand the difference between something that can learn on it's own like an ANN and something wherein the programmer must input all the rules for the AI to follow manually like for example most text-based games I've seen before.
What games do you recommend me looking into? I've never seen a text-based game that leveraged machine learning, but maybe there were some I never knew about.
A lot of older games used this - there's a Hitchhikers Guide one my father harps on about.
Agree about AI being misunderstood, but... DDB has a bad track record in terms of software development #homebrewer
I play this game because of the interactions I have with the other players / dms at the table. Having a computer just run me through a game seems pretty unappealing. Unless it's a complete mess, then I could see it as a fun romp for a laugh.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
1) I think framing the issue as a zero-sum game between "increasing the capabilities of AI" and "reducing the social elements of DnD" is a false dichotomy to begin with. The former does not require the latter.
2) While training AI can involve a lot of ethical quandaries, in this case there are none - it would have to be trained entirely on games run through the DDB service or any partnering VTTs, which all of us would be explicitly agreeing to by using those. It's not like they'd be able to monitor meatspace games even if they wanted to.
These are all great ideas, however, going in this direction does open players up to cheaters, bots 'and the whole nine'.
Which is the reason I refuse to play massive multi-player or online video-type games at all.
The live human element in tabletop games is essential to Dungeons & Dragons as mentioned earlier in this thread.
The experience I am striving for is playing D&D with real people with the option of making real friends.
Video games are a far departure from that. Trust me, I know from experience.
I feel like this would end up being the worst of both worlds between TTRPGs and video games.
You have the limited options of a video game, where the devs can only account for so many factors from the players. Even games where you have choices on things, tend to limit you to a few. On the other hand, you don't have the full on production value of a game like Baldur's gate or Divinity 2 Original Sin.
I wonder how it would even train full sessions. A lot of people playing online use discord or other programs to chat for example, how do you process that? Or do you try and feed it without the context from voice chat? How do you get it to incorporate homebrew or house rules?
I could see an AI handling the pure war game, RAW aspects of the game, but struggling with even things like adapting the AI of how say, a swarm of goblins approach fighting the party as opposed to a battalion of knights. Let alone aberrations and dragons and fae etc. Just making the fights feel dynamic and not like you're plying out a random encounter in a turn based JRPG on a grid feels like it may be difficult for an AI to authentically reproduce.
If this actually happens, then good luck to them. I'm skeptical but I'm not necessarily against the idea as long as people have to opt in to having their sessions fed into the AI.
If there's such a dungeon master shortage in D&D5, stand up and run some campaigns.
J
Great Wyrm Moonstone Dungeon Master
The time of the ORC has come. No OGL without irrevocability; no OGL with 'authorized version' language. #openDND
Practice, practice, practice • Respect the rules; don't memorize them • Be merciless, not cruel • Don't let the dice run the game for you
I just saw the slightly negative comment on Gloomhaven and got a bit mad. But otherwise, I agree with what you said. Gloomhaven pretty much satisfies my 'need' for the combat bit of DnD and other TTRPGs, so I mostly play DnD to roleplay. I have a sneaking suspicion that A.I dms will never really be able to be as good as an actual DM, but who knows.
N/A
*The best AI DM online at the moment sucks ngl*
Frequent Eladrin || They/Them, but accept all pronouns
Luz Noceda would like to remind you that you're worth loving!
Welllll.... Not to be that guy, but you already agreed to give them access to any of your usage data for any Website, Game, or Service they offer, just by posting here or using any of their online sites/tools. We all did.
It's an opt-out system --
we opt to play D&D on a VTT that is out in the community. ;)
J
Great Wyrm Moonstone Dungeon Master
The time of the ORC has come. No OGL without irrevocability; no OGL with 'authorized version' language. #openDND
Practice, practice, practice • Respect the rules; don't memorize them • Be merciless, not cruel • Don't let the dice run the game for you
WOTC IA would not be able to run as a DM in a game session under the current rules. I have a good understanding of this from what I used to do. BG3 in it's turn assisted mode needs a really large install on a very expensive computer to do what it does and it is limited to a very restricted rule set, not the current rules.
WOTC and HASBRO are aiming at a presentation layer running in a web browser on a computer, tablet or cellphone like the current D&D Beyond character sheet.
That is never going to happen within the next 5 to 10 years on a computer or really powerful laptop in a web browser and certainly never on a cell phone in my lifetime even if some clueless executive at HASBRO thinks otherwise with the current rules set.
The 6th Edition of D&D would have to be gutted massively from the current 5th edition to get it to work. Think the difference between pong and BG3 where 6th Edition would be pong...
edit added " from the current 5th edition "
Hell no! No AI at all. Say no to the $30 fee too. I can barely afford the tier I do have, and most of my stuff is homebrew.
D&D was made popular on its accessibility. Anything that detracts from that will kill the game, allowing competitors to step up.
Hasbros/WotC... you're f-ing idiots if you are even considering this. I didn't cancel during the OGL hoopla, but if you go this route, I will.
I'm playing homebrew with live people, whether you like it or not.
Period.
I am an experienced player (approximately 40 years) who is always looking for a group.
I would agree to some small degree. I think that an AI DM could have some use. It might be good for a very niche use, even if they don't generally replace human DMs. In a situation where there is no way in which you could have a human DM or you want a solo experience, for example. However, there are two main reasons I think they will never actually replace hman DMs:
1. A large part of D&D for many people is the human interaction and the shared experience created. It's not just about telling a story, after all I could write a book if that's what I wanted. Nor about the mechanical gameplay, I could play a regular board game if that's all I wanted. Yes the fusion of these two is part of what makes D&D so awesom, but so is the fact that it creates a shared experience between real human people.
2. This is an even bigger issue: DMs can give very tailored experiences. There's so many ways in which a game can be subtly manipulated so give thousands of different possible experiences. An AI can have "settings" like "comedic," "dark," "heavy on RP," but will never be able to do what people can, at least with current technologies. AIs take absolutely gargantuan amounts of data to learn compared to humans, so it simply isn't possible for them to learn about the preferred play style of the player(s) in a feasible amount of time.
As someone who works on AI, this is simply not true. There is one critical difference, and one that makes massive differences. An ordinary video game uses conventional algorithms. That is, the programmer creates basically a set of "rules," which the program follows. If you do X, the game responds with Y. However, in a machine learning system, the programmer creates a structure to learn from data (usually a neural net but their are other approaches like rule based learning to decision trees), which allows it to find patters in the data. In can then, knowing this pattern, find the same pattern in thing it hasn't seen before. Oversimplification, but that's basically how it works.
So in a video game, everything that can be done has to be programmed by the creators. The player can only affect the plot to a certain degree because the developers have to have created all the possible endings and so on. The players can also only take certain actions the the developers integrated into their code. The list goes on.
Using a machine learning system on the other hand allows the program to react in new ways that were never envisioned by the developers. If you do something new, it can significantly change the plot based on your actions. RP and dialouge can be real, as demonstrated by chatGPT and others before it. You can find new solutions to problems and make your own decisions. With a sufficiently advanced ML system, there may not even be a detectable difference between an AI and a human (which has already been shown in published research in some other contexts such as conversation and composition). The main difference is not the actual game itself but the human interaction that is part of it.
I am an average mathematics enjoyer.
>Extended Signature<
It's true, but then again it hasn't got even close to the amount of research as other ML applications, whether that be image recognition or even AI legal judges. I don't think that now is the time for an AI DM, but I don't think its impossible. I don't think it would be as good as real D&D, but it could be a lot better than AI dungeon.
I am an average mathematics enjoyer.
>Extended Signature<
I can see a DM assistant.
A program that would help the DM set up things.
Like monster choices. The Dm could input the players levels and classes and the Dm-Ast would list out all the possibilities. The DM would then just click on what they want and the stats would be posted and kept track of.
How about a list of normal NPC's to populate a town. Just input the towns size and shops can be listed and proprietors randomly named.
All of this could be based on settings and town sizes.
We do not need infinitely random mapping. A growing list of maps for each town size. The Dm fills in the details. The same for underground settings, Castles and other lairs. I have found that truly random map makers jest get it all wrong.
I'm with Yurei on that one. An AI DM becomes a video game.
The whole point of having a DM is to handle whatever the players may come up with where a computer game is built to manage and thus limit the outcome in its own boundaries so that it can react to what has been predicted by the programmers. Remove the players creativity and you've removed the purpose of the game itself...
If some of you have played Bloodbowl 1 and 2, you have probably realized how terrible the AI is and they have been working on it for years. I could even argue that the 95 dos game had actually a better AI. Compare this to what D&D would require, it is not even thinkable if you ask me.
If you'd want to build an AI, the first thing you'd want to provide is how to handle the Rules but also every possible scenario how the rule may be affected by other variable that may factor in. That alone would be a piece of work and it is not always obvious as there are many question that ends up being answered on twitter or else. That alone could provide a great tool for a GM.
The second thing that comes to mind building an AI, is if you're planning on working in specific settings such as The Forgotten Realm, being able to give the lore of every region including cities would be in itself a great piece of art. That again, could be a great tool for a GM. The party decides to go to Longsaddle suddenly and you're wondering what you're supposed to describe, type it in on the AI and you'd have something in hand to make it closer to the lore.
This is a fundamental misunderstanding of what AI is capable of. What makes AI both so useful and so revolutionary is that it can react to situations to programmers never thought of or hardwired into the system. It's called "machine learning" for a reason. The AI literally learns for itself how to do all this, usually by looking at data (though there is also reinforcement learning which still in many ways an emerging technology). For an AI to work, all you need to do is give it large amounts of data and give it the simple equations that govern back propagation (they're basically just the chain rule from calculus). If you don't believe me, check out chatGPT, word2vecc, midjourney, or even alphazero: they all can handle basically anything you throw at them, even weird and strange things that the developers would have never imagined yo using it for.
You're comparing the general concept "AI" to something that could in the loosest definition be considered an AI but if far weaker and more dependent on human inputs then the current AIs that exist. There aren't any really great AIs in the field of role playing mainly because it's not that important in context of all the other things AIs are used for. There just hasn't been a lot of serious research on this.
But even what has been made goes beyond this generalization of AI you are making. Comparing AIs to gloomhaven is an unfair comparison because gloomhaven is both a board game which cannot access the power computing power of a computer, and thus could never actually have an AI (ANNs take crazy amounts of computing power to run), AND never was advertised as being "an AI RPG." If we look at actual examples of AI RPGs, one of the best examples is AI dungeon. Sure, it's not great and not D&D, but it's definitely not a video game, at least not in the way as in it is not like D&D. It can react to most player decisions and change the plot accordingly. It doesn't just follow input programmed by the developer and can actually tell a cohesive story in which the player has the driving force and can significantly alter the plot in any number of ways, just like D&D. It's not meant to be D&D and has a ton of room for improvement, but I think it shows that an AI DM is possible. With a lot more research, effort, and funding I'm sure an AI DM that can act fairly similarly to a normal DM is possible.
I am an average mathematics enjoyer.
>Extended Signature<
I encourage people to look into those old epic text-based video games and tell me that a digital DM is different.
Frequent Eladrin || They/Them, but accept all pronouns
Luz Noceda would like to remind you that you're worth loving!
I will, but I can tell you before even starting that AI is a rapidly evolving field which has seen gargantuan innovation even in the last few years. AIs like StableDiffusion, ChatGPT, and AlphaZero would have been thought impossible in 2010 even. Plus with rapid advances just in the computing power, deep neural networks were basically just made possible in the 2000s. I think the term "AI" is often used for something that is not really what AI is capable of, leading people to misunderstand the difference between something that can learn on it's own like an ANN and something wherein the programmer must input all the rules for the AI to follow manually like for example most text-based games I've seen before.
What games do you recommend me looking into? I've never seen a text-based game that leveraged machine learning, but maybe there were some I never knew about.
I am an average mathematics enjoyer.
>Extended Signature<
A lot of older games used this - there's a Hitchhikers Guide one my father harps on about.
Agree about AI being misunderstood, but... DDB has a bad track record in terms of software development #homebrewer
Frequent Eladrin || They/Them, but accept all pronouns
Luz Noceda would like to remind you that you're worth loving!
Anyone remembers a text based game called 'Zork'?
I play this game because of the interactions I have with the other players / dms at the table. Having a computer just run me through a game seems pretty unappealing. Unless it's a complete mess, then I could see it as a fun romp for a laugh.