Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Worlds first ever RPG with 'dungeon master ai' and 'story engine' in the works

agamennagamenn Member UncommonPosts: 67

Developers want to create a human like AI dungeon master in the game that will react to what you do in the game, they want to create a real 'do what you want to do, go where you want to go' type game based on tabletop D&D type adventures.
AlBQuirkyblamo2000AmatheultimateduckUngood
«1

Comments

  • kitaradkitarad Member EpicPosts: 5,522
    Tagging this must watch it later.
    AlBQuirky

  • VrikaVrika Member EpicPosts: 6,088
    I didn't watch the video because it's too long, but this won't work. AI generated content isn't close enough to handcrafted content yet.

    They may be able to make another roguelike game that ends up good, but that's the extend of what todays AI tech can make. 
    blamo2000
     
  • agamennagamenn Member UncommonPosts: 67
    Vrika said:
    I didn't watch the video because it's too long, but this won't work. AI generated content isn't close enough to handcrafted content yet.

    They may be able to make another roguelike game that ends up good, but that's the extend of what todays AI tech can make. 


    They have a very good programmer (Julian Lafey) who has been working on the engine for this for years now.

    Roguelike comparison is good, but this will be a rogulike with story and persistence.
  • AlBQuirkyAlBQuirky Member EpicPosts: 5,862
    edited November 2019
    Marked for "watch later", too. An hour and a half of Q&A intrigues me ;)

    It will be interesting to see how these folks do. The biggest thing I missed from tabletop D&D to video games was the reaction the DM had to our actions. Video games are just canned stories with few, if any branches. Too few of them "evolve" much, if at all and I always feel "forced" into following the storylines, no matter what choices I make.
    Palebane

    - Al

    Personally the only modern MMORPG trend that annoys me is the idea that MMOs need to be designed in a way to attract people who don't actually like MMOs. Which to me makes about as much sense as someone trying to figure out a way to get vegetarians to eat at their steakhouse.
    - FARGIN_WAR


  • blamo2000blamo2000 Member RarePosts: 1,007
    A minute long trailer with gameplay would be much better than a movie length video of people talking to sell this (in my opinion).  

    This sounds interesting, and as long as it ticks all my boxes and doesn't do anything egregious (like have like "decks" build around a card games or any nonsense like that), I will check it out.

    But like Vrika above said - I think scripting and a human putting effort into thinking out what or how a person or group could respond gives much better results in an rpg than AI.  But, without devs pursuing the possibility of AI and what it can do, or pushing towards what it hopefully will do in the future, we will never get there.  
    AlBQuirkyMaurgrimMendelNarug
  • SovrathSovrath Member LegendaryPosts: 27,931
    Vrika said:
    I didn't watch the video because it's too long, but this won't work. AI generated content isn't close enough to handcrafted content yet.

    They may be able to make another roguelike game that ends up good, but that's the extend of what todays AI tech can make. 
    hmmm maybe.

    It really depends how much granularity you are are looking for. There isn't going to be "Star Trek the Next Generation" levels of Holodeck adventure building but I can imagine that there are ways this can be pulled off in a reasonable way.
    AlBQuirky



  • SovrathSovrath Member LegendaryPosts: 27,931
    blamo2000 said:
    But, without devs pursuing the possibility of AI and what it can do, or pushing towards what it hopefully will do in the future, we will never get there.  
    pretty much this. We need people to be pushing the envelope of what can be done.
    AlBQuirkyblamo2000Palebane



  • MaurgrimMaurgrim Member RarePosts: 1,271
    edited November 2019
    They talk about it didnt see any ingame footage which is 10000000000000 times more interesting than just doing selling pitches on a 1.5 hour Q&A video.
    Kyleran
  • agamennagamenn Member UncommonPosts: 67
    Maurgrim said:
    They talk about it didnt see any ingame footage which is 10000000000000 times more interesting than just doing selling pitches on a 1.5 hour Q&A video.
    They are early in production, but we get an idea of what they want to make.
  • MendelMendel Member EpicPosts: 3,672
    Sounds like these people are willing to take some risks.  But, the fact is that presently they "want" to do this, not "have" done it.  Whether or not they can accomplish a "DM AI" seems like a very powerful step forward for the genre.  Fail and they learn their limits.  Succeed and we have a playable game.  I equate this with "conceptual game design", much like conceptual art for now.  I am curious enough to watch the full video when I have time to spare.

    A quick prediction.  The real challenge this effort will face will be determining how to abstract user actions.  Fundamentally, any game needs to know what inputs to expect and how to respond to those inputs.  How do they intend to abstract an action they haven't anticipated and coded?



    AlBQuirky

    Logic, my dear, merely enables one to be wrong with great authority.

  • AmatheAmathe Member LegendaryPosts: 6,686
    I would certainly give this a try.

    EQ1, EQ2, SWG, SWTOR, GW, GW2 CoH, CoV, FFXI, WoW, CO, War,TSW and a slew of free trials and beta tests

  • SovrathSovrath Member LegendaryPosts: 27,931
    Mendel said:
    Sounds like these people are willing to take some risks.  But, the fact is that presently they "want" to do this, not "have" done it.  Whether or not they can accomplish a "DM AI" seems like a very powerful step forward for the genre.  Fail and they learn their limits.  Succeed and we have a playable game.  I equate this with "conceptual game design", much like conceptual art for now.  I am curious enough to watch the full video when I have time to spare.

    A quick prediction.  The real challenge this effort will face will be determining how to abstract user actions.  Fundamentally, any game needs to know what inputs to expect and how to respond to those inputs.  How do they intend to abstract an action they haven't anticipated and coded?



    I haven't had a chance to watch the whole video but I think it depends on their scope. There are definitely things they can plan for. However, in a real "live" pen and paper game, players can come up with a huge amount of variation, even things that go beyond the scope of the adventure.

    It's my guess that they are talking more about "I use a fireball/I go down this hall over that hall/I ingest this potion at this time", etc.

    Over players saying "I find a pickaxe and I try to burrow through the wall."
    AlBQuirky



  • cheyanecheyane Member EpicPosts: 6,705
    I always wished to play in a game like the one in the book Dream Park by Larry Niven and Steven Barnes. I really enjoyed that book and often wondered how great it would be to play something like that. I don't want the horrific things that happened but the idea definitely made me imagine things.
    blueturtle13
    image
  • agamennagamenn Member UncommonPosts: 67
    edited November 2019
    "How do they intend to abstract an action they haven't anticipated and coded?"

    In an interview that Julian gave, he said he wants to put in neural network and machine learning to create the AI.

    AI would essentialy take action based on what the developers made it learn, and as it is 'dungeon master' the AI would be only limited to certain things it can do.

    I guess by 'dungeon mater' they mean the AI would create stories for you, basicaly quests - you go into a town and the AI creates a story for you in that town..... you go into a cave instead of a town and the AI creates a story for you in the cave.
    Mendel
  • VrikaVrika Member EpicPosts: 6,088
    edited November 2019
    agamenn said:
    "How do they intend to abstract an action they haven't anticipated and coded?"

    In an interview that Julian gave, he said he wants to put in neural network and machine learning to create the AI.

    AI would essentialy take action based on what the developers made it learn, and as it is 'dungeon master' the AI would be only limited to certain things it can do.

    I guess by 'dungeon mater' they mean the AI would create stories for you, basicaly quests - you go into a town and the AI creates a story for you in that town..... you go into a cave instead of a town and the AI creates a story for you in the cave.
    The problems with that are:
     -There is no way to automatically measure a story. The AI doesn't learn anything unless you've got a human who rates the story
     -Machine learning is good when your aim is to have the machine arrive in same or similar solution in similar situation. For creating stories it doesn't work that well because it tends to reach solutions like "Most towers have an imprisoned princess" rather than "Having a story about princess imprisoned in a tower would be good"

    Learning AI could work well for placing things like traps and monsters, where you can make automatic ways to measure how well party is doing, and the player is happy enough if the orc camp has 10 slightly different groups of orcs in it. It's not good for stories that can't be automatically measured and the player expects to be different from each other.
    MendelPalebane
     
  • MendelMendel Member EpicPosts: 3,672
    agamenn said:
    "How do they intend to abstract an action they haven't anticipated and coded?"

    In an interview that Julian gave, he said he wants to put in neural network and machine learning to create the AI.

    AI would essentialy take action based on what the developers made it learn, and as it is 'dungeon master' the AI would be only limited to certain things it can do.

    I guess by 'dungeon mater' they mean the AI would create stories for you, basicaly quests - you go into a town and the AI creates a story for you in that town..... you go into a cave instead of a town and the AI creates a story for you in the cave.
    My initial point was that players are inventive, computers aren't.  When faced with a chest that they can't pick, players are likely to try to take the entire chest, or burn the chest, or pour acid in the lock.  What exactly is the UI icon to suggest pouring acid in the lock?  How is that icon distinct from one for pouring water in a lock or pouring acid over the hinges?  That's what is going to be difficult.  Words can be precise and flexible, graphic icons not so much.

    In order for a computer to react to an action, the player needs to first input that action.  Without the ability to input a specific action, even the best computer learning machines can't learn.



    AlBQuirkyPalebane

    Logic, my dear, merely enables one to be wrong with great authority.

  • PalebanePalebane Member RarePosts: 3,916
    edited November 2019
    Would be cool if they could afford to hire some top notch human DMs, imo. Imagination is key.

    Vault-Tec analysts have concluded that the odds of worldwide nuclear armaggeddon this decade are 17,143,762... to 1.

  • SovrathSovrath Member LegendaryPosts: 27,931
    Mendel said:
    agamenn said:
    "How do they intend to abstract an action they haven't anticipated and coded?"

    In an interview that Julian gave, he said he wants to put in neural network and machine learning to create the AI.

    AI would essentialy take action based on what the developers made it learn, and as it is 'dungeon master' the AI would be only limited to certain things it can do.

    I guess by 'dungeon mater' they mean the AI would create stories for you, basicaly quests - you go into a town and the AI creates a story for you in that town..... you go into a cave instead of a town and the AI creates a story for you in the cave.
    My initial point was that players are inventive, computers aren't.  When faced with a chest that they can't pick, players are likely to try to take the entire chest, or burn the chest, or pour acid in the lock.  What exactly is the UI icon to suggest pouring acid in the lock?  How is that icon distinct from one for pouring water in a lock or pouring acid over the hinges?  That's what is going to be difficult.  Words can be precise and flexible, graphic icons not so much.

    In order for a computer to react to an action, the player needs to first input that action.  Without the ability to input a specific action, even the best computer learning machines can't learn.



    But that's the thing, that's not really hard. As long as it's within the context of what the object can reasonably receive as actions.

    The player comes to the chest. He "can" take it. As long as he/she can carry it. Done. Take it into town. That's not hard to do.

    As far as the lock? Assign the "pick/break/acid/whatever" and then some outcomes with die rolls.

    My guess is that all the actions a player might do will be there. Not hard. Now, "other things" such as fill the lock with feathers or any number of creative things, won't be there. But that's for players to temper their expectations.
    AlBQuirky



  • EladiEladi Member UncommonPosts: 1,121
    Impossible venture atm, yes they could do some nifty things whit AI, but cant come near the way a real player could react to a situation. for that you need AI that can almost mimic the human state of mind whitin the moment a human player reacts to a situation.  This a AI will never ever be able to do even if it had direct acces to all the world knowlidge and a way to near instandly process sutch data.  the only way it will ever come close is when we humans are able to make a direct copy of our brain workings into a mechanical version. (brain in a jar..) 

    IF we would reach this point of game development in our lifetime we would also have 100% VR and eliminated death more or less.  

    Its all on the same scale as 42 
  • AlBQuirkyAlBQuirky Member EpicPosts: 5,862
    For those wanting a tl/dr version, I just watched this video not realizing it was about the same game.


    Relampago

    - Al

    Personally the only modern MMORPG trend that annoys me is the idea that MMOs need to be designed in a way to attract people who don't actually like MMOs. Which to me makes about as much sense as someone trying to figure out a way to get vegetarians to eat at their steakhouse.
    - FARGIN_WAR


  • agamennagamenn Member UncommonPosts: 67
    Eladi said:
    Impossible venture atm, yes they could do some nifty things whit AI, but cant come near the way a real player could react to a situation. for that you need AI that can almost mimic the human state of mind whitin the moment a human player reacts to a situation.  This a AI will never ever be able to do even if it had direct acces to all the world knowlidge and a way to near instandly process sutch data.  the only way it will ever come close is when we humans are able to make a direct copy of our brain workings into a mechanical version. (brain in a jar..) 

    IF we would reach this point of game development in our lifetime we would also have 100% VR and eliminated death more or less.  

    Its all on the same scale as 42 


    We have to wait and see :P
  • agamennagamenn Member UncommonPosts: 67
    edited November 2019
    i.imgur.com/WCpAzoH.jpg
  • skeaserskeaser Member UncommonPosts: 4,051
    AI worked well for EQ Next and storybricks...
    Sig so that badges don't eat my posts.


  • vegetableoilvegetableoil Member UncommonPosts: 507
    I'ts called Procedural generated games. ring a bell?
  • anemoanemo Member RarePosts: 1,865
    I'ts called Procedural generated games. ring a bell?
    Most procedurally generated games only really have the ability to "generate" at the start of world generation.   The ability to realize that it needs to increase the number of "crafting encounters" because the player is starting to run out is pretty rare.  NPCs realizing that the player likes to get in on "This type of job, so we should get them in on this one",  is also pretty rare for procedurally generated worlds.

    The closest adaptive systems we have in gaming right now are related to adaptive matchmaking for PvP.   Where the matchmaker takes any data available (skill levels, prefered class types, types of players the player has worked with well) and aims to get a targeted win rates.

    So it would make sense that a dev somewhere would want to try and play with such a system brought into the PvE space.

    Practice doesn't make perfect, practice makes permanent.

    "At one point technology meant making tech that could get to the moon, now it means making tech that could get you a taxi."

Sign In or Register to comment.