SC's Subsumption #nobullshit

MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
edited September 25 in Star Citizen
For those interested in the actual game and not the classic forum drama about it ~~

A short and direct video that explains Subsumption, one of the main pitches and a pillar of SC and SQ42, this deeply relates to AI, economy, and missions, it's a system releasing its first iteration in the next 3.0 update. The AI of SC is more like The Sims, with a system "whose purpose is to define behavior based on sensory input, goals and rules",  unlike the NPCs on rails continuously it in a loop that never changes you normally find in MMO's.



Wouldn't you agree more MMO's should have a more dynamic "The Sims" like AI for NPCs?

edit - In another more recent bit but related and a good example of how it drives some game systems, this on the ATC (Air Traffic Controller) where the NPCs provide that service themselves:

Post edited by MaxBacon on
Octagon7711ExcessionRhoklawGdemamiGruugAzaron_Nightblade
«13

Comments

  • Slapshot1188Slapshot1188 Boca Raton, FLMember EpicPosts: 7,186
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    TalonsinVrikaScotchUpbcbullyExcessionKefoMrMelGibsonKyleranNycteliosBeansnBreadand 3 others.

    "I should point out that no other company has shipped out a beta on a disc before this." - Official Mortal Online Lead Community Moderator

    Starvault's reponse to criticism related to having a handful of players as the official "test" team for a supposed MMO: "We've just have another 10ish folk kind enough to voulenteer added tot the test team" (SIC) This explains much about the state of the game :-)

  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    It's a design ofc, but most do not even try to do one MMO with one different AI system, they just go with the classic design the others do. This has been talked about since the day this game was pitched, this took quite a long time until the AI starting to translate from design to implementation.

    Though I meant the question more related to "the sims" like AI, I'll edit that bit.
    ExcessionGdemami
  • MadFrenchieMadFrenchie Nashville, TNMember EpicPosts: 2,386
    edited September 17
    MaxBacon said:
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    It's a design ofc, but most do not even try to do one MMO with one different AI system, they just go with the classic design the others do. This has been talked about since the day this game was pitched, this took quite a long time until the AI starting to translate from design to implementation.

    Though I meant the question more related to "the sims" like AI, I'll edit that bit.
    I think @Slapshot1188 was implying there's an underlying reason MMORPGs haven't attempted this sort of thing.

    Don't get me wrong Max, I'm hugely disappointed in the lack of AI progression in MMORPGs and would love to see leaps in that department, but we all know it's a tough and tricky road that requires massive amounts of resources to pull off.  If SC releases in a stable PU state with this kind of system being successfully implemented, I'll be a huge fan for that single feature.  But it's easier said than done, and by done, I think it's fair for me to say "we" mean included as a system within a working MMORPG game as a whole.
    Post edited by MadFrenchie on
    MrMelGibsonExcession

    image
  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    I think @Slapshot1188 was implying there's an underlying reason MMORPGs haven't attempted this sort of thing.

    Don't get me wrong Max, I'm hugely disappointed in the lack of AI progression in MMORPGs and would love to see leaps in that department, but we all know it's a tough and tricky road that requires massive amounts of resources to pull off.  If SC releases in a stable PU state with this kind of system being successfully implemented, I'll be a huge fan for that single feature.  But it's easier said than done, and by done, I think it's fair for me to say "we" mean included as a system within a working MMORPG game as a whole.
    I can imagine the reasons.

    It's resource-costly to simulate AI like that. It's not really "can it be done?", it's more it's by nature complex what adds up to the fact it needs to play out in the MP environment of the game.

    The more costly the game-servers get to the run, fewer profits there are, that alone would be a reason to why avoid simulation (AI/Physics) of relevant complexity.
    Gdemami
  • Octagon7711Octagon7711 Chicago, ILMember EpicPosts: 5,711
    I think the problem is SC is still in a concept heavy phase.  People want to see more finished concepts at this point. But this is alpha which is known for being concept heavy then shifting to a focus of working on a finished product.  Because CIG makes no promises, it's hard to tell if this is still the start middle or near the end of alpha and how long each phase is supposed to last.  

    "Change is the only constant."

  • laxielaxie UK - Leamington SpaMember RarePosts: 800
    I like the idea of a "meta server" that holds broad information, and then creates instances based on the current meta state. Even SWG had glimpses of this back in the day, where the cities were spawning either imperial or rebel NPCs, based on who was in control of the territory. This doesn't even require that much processing power, as checking what to spawn based on criteria is not that complex. You can get a good sense of player impact this way, while keeping the resource cost minimal.

    In terms of multi-layered NPCs with complex decision-making, I think this doesn't really make sense in the context of Star Citizen. The "meta server" approach makes perfect sense, but making decisions on the fly is a whole different story. The resources to pull that off would be quite extensive, especially if it is able to respond to a variety of situations.

    I'm saying it doesn't make sense because of the relatively low player numbers within an instance. To run a detailed simulation of a space station's inhabitants for 50 people in an instance would be very costly. I think this makes more sense for non-instanced games, where you are getting a bit more bang for your buck. Having John, the fruit seller in Station X, instanced over 50 different variations of Station X is fairly inefficient, especially if he is to make complex unique decisions (based on what the players do) in each one of them.
    Gdemami
  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    I think the problem is SC is still in a concept heavy phase.  People want to see more finished concepts at this point. But this is alpha which is known for being concept heavy then shifting to a focus of working on a finished product.  Because CIG makes no promises, it's hard to tell if this is still the start middle or near the end of alpha and how long each phase is supposed to last.  
    Specific to subsumption though it's already in for 3.0 so what they talk about is way more fleshed out than what was just one idea back then. This obviously by iterations and the first release is the basic version of what the AI meta is.

    laxie said:
    I like the idea of a "meta server" that holds broad information, and then creates instances based on the current meta state. Even SWG had glimpses of this back in the day, where the cities were spawning either imperial or rebel NPCs, based on who was in control of the territory. This doesn't even require that much processing power, as checking what to spawn based on criteria is not that complex. You can get a good sense of player impact this way, while keeping the resource cost minimal.

    In terms of multi-layered NPCs with complex decision-making, I think this doesn't really make sense in the context of Star Citizen. The "meta server" approach makes perfect sense, but making decisions on the fly is a whole different story. The resources to pull that off would be quite extensive, especially if it is able to respond to a variety of situations.

    I'm saying it doesn't make sense because of the relatively low player numbers within an instance. To run a detailed simulation of a space station's inhabitants for 50 people in an instance would be very costly. I think this makes more sense for non-instanced games, where you are getting a bit more bang for your buck. Having John, the fruit seller in Station X, instanced over 50 different variations of Station X is fairly inefficient, especially if he is to make complex unique decisions (based on what the players do) in each one of them.
    I would believe they add this on SC because multi-crew ships, even if you have a player crew you'll still need AI crew and that AI is stated to be just like SQ42, so your crew will need that sort of simulation to react all sorts of situations.

    What you are talking about I think was that old design when the game was fully instanced the "Universe Servers" to handle AI, interacting with all the instances. They seem to have been changing that to localize processing of areas of space per server instead of instancing the same place multiple times, I think we may end up with multiple "master regions" instead of copies of the same instance, that's to be seen.
    ExcessionGdemami
  • hfztthfztt GlostrupMember RarePosts: 1,242
    Funny story. Remember the Mass Effect Andromeda lol's regarding face animation? Those animations was build on a similar system. The more complex the calculation, the harder to test, the more likely to result in some really strange shit.

    But props to them for trying. Just be realistic about it. There is a reason many other games have started work on similar systems, yet ended up with something more predictable.
  • btdtbtdt Member UncommonPosts: 98
    As my brother always used to say, anything can be done if you throw enough money at it... the question is, whether it worth it?


    Ultimately the game has to be worth playing, everything else is just window dressing.  Sure the graphics are stunning... but the game is boring as hell.  Sure the combat feels glorious... but the game is totally unbalanced.  Sure the story is outstanding... but there's no replayablity.

    This stuff is all window dressing to disguise the fact that the game itself, lacks something at it's core.  Back in the day you played games for hours on end, not because of the quality of the graphics, dynamic events, et al... but because it was entertaining.  Some of the simplest designs and implementations are far more engaging than the most detailed designs.  

    As my professor used to say, God is in the design, the Devil is in the details.  Details don't make the design and the design doesn't make the details.  It's a careful balance between the two.  Your brain filters out a lot of the details, to get to the root design.  Spending an inordinate amount of time, energy, and money on something the brain will treat as background noise, doesn't make it better.

    Hence why some games don't seem to change the formula up... they already know you're going to filter it out in short order.
  • XiaokiXiaoki White Pigeon, MIMember RarePosts: 2,872
    MaxBacon said:
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    It's a design ofc, but most do not even try to do one MMO with one different AI system, they just go with the classic design the others do. This has been talked about since the day this game was pitched, this took quite a long time until the AI starting to translate from design to implementation.

    Though I meant the question more related to "the sims" like AI, I'll edit that bit.
    The reason why MMOs go with standard AI is the very nature of MMOs: the internet.

    AI has to be calculated server side because if its calculated client side then it can(and will) be easily hacked to pieces.

    Now, with anything thats done server side you cant calculate it once then send it out to everyone, no, you have to calculate it separately for everyone.

    Lets say 10 people are looking at 1 mob. That means the server has to calculate the AI for that 1 mob 10 times. If you have 10 people looking at 10 mobs then the server has to calculate the AI for those 10 mobs 100 times.

    So, it has never been a matter of companies not wanting or being unable to make better AI, it has always been a matter of having to send millions complex AI calculations to hundreds of thousands of people over the internet every second.

    A couple MMOs have tried complex AI for mobs. Namely, Project Titan by Blizzard and Everquest Next by SOE. Now, who can tell me what happened to them?
    Arglebargle
  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    hfztt said:
    Funny story. Remember the Mass Effect Andromeda lol's regarding face animation? Those animations was build on a similar system. The more complex the calculation, the harder to test, the more likely to result in some really strange shit.

    But props to them for trying. Just be realistic about it. There is a reason many other games have started work on similar systems, yet ended up with something more predictable.
    AI for the faces? O___o But yeah the nature of it is that the more simulation the harder it is to control the output, it requires a massive amount of logic.

    But I think we've been having very VERY basic AI in MMO's, and that AI is very core to the game and its content, seeing GW2 in that aspect shown that (even though fully scripted in a loop) AI meant to create a believable dynamic world created one of the best PvE Gameworld experience I've had in MMO's.
  • hfztthfztt GlostrupMember RarePosts: 1,242
    Xiaoki said:

    Lets say 10 people are looking at 1 mob. That means the server has to calculate the AI for that 1 mob 10 times. If you have 10 people looking at 10 mobs then the server has to calculate the AI for those 10 mobs 100 times.

    I do not understand or agree with your argument here. As I see it the AI still only have to do one calculation on the server, just taking the 10 players into account, getting one result and sending it to all players, just the same as with a simple AI.
  • Octagon7711Octagon7711 Chicago, ILMember EpicPosts: 5,711
    edited October 2
    Don't worry, what ever they come up with I'm sure it will get reworked a few more times.  
    Post edited by Octagon7711 on
    Arglebargle

    "Change is the only constant."

  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    edited September 17
    btdt said:
    As my brother always used to say, anything can be done if you throw enough money at it... the question is, whether it worth it?


    Ultimately the game has to be worth playing, everything else is just window dressing.  Sure the graphics are stunning... but the game is boring as hell.  Sure the combat feels glorious... but the game is totally unbalanced.  Sure the story is outstanding... but there's no replayablity.

    This stuff is all window dressing to disguise the fact that the game itself, lacks something at it's core.  Back in the day you played games for hours on end, not because of the quality of the graphics, dynamic events, et al... but because it was entertaining.  Some of the simplest designs and implementations are far more engaging than the most detailed designs.  

    As my professor used to say, God is in the design, the Devil is in the details.  Details don't make the design and the design doesn't make the details.  It's a careful balance between the two.  Your brain filters out a lot of the details, to get to the root design.  Spending an inordinate amount of time, energy, and money on something the brain will treat as background noise, doesn't make it better.

    Hence why some games don't seem to change the formula up... they already know you're going to filter it out in short order.
    I think it is worth it.

    I'll use the example of Guild Wars 2 here, one of the selling points of the game was the claimed dynamic world/events, and that system runs under a fully scripted loop that never changes.

    Even so, it was the best approach I've seen on it in one MMO to date, but then...

    What about turning those scripts in a loop into actual AI simulation so say fights, invasions, events, etc... are driven by simulation leading to dynamic chains of events triggered by that simulation and/or players?

    I would find that amazing because then it would not be the same thing in a loop until the end of the times and there would exist an actual dynamic to it that would make it feel like far more like a living world, instead of make-believe.

    Would it cost more to have that simulation on servers? Sure. Would it be more complex for the developers to work in the colossus of logic necessary to ensure the simulation plays like the script would? Sure. But I think that would create one experience that adds far more replayability and longevity to the gameplay per area.
    Post edited by MaxBacon on
  • hfztthfztt GlostrupMember RarePosts: 1,242
    edited September 17
    MaxBacon said:
    hfztt said:
    Funny story. Remember the Mass Effect Andromeda lol's regarding face animation? Those animations was build on a similar system. The more complex the calculation, the harder to test, the more likely to result in some really strange shit.

    But props to them for trying. Just be realistic about it. There is a reason many other games have started work on similar systems, yet ended up with something more predictable.
    AI for the faces? O___o 
    Yup. The point was to not have to hand animate everything (comparable to hand script missions). So they had code that calculated how to animate the character based on the situation, environment, previous story choices and nature of the dialog (again comparable to the mission setup seen here).

    So yeah. Pretty comparable actually.

    Of course they where up against the fact that reading faces is a thing we humans are very good at, so small deviations seem huge to us. We are much less likely to spot "silly faces" in a mission configuration as we do not have a natural frame of reference for that.
    Post edited by hfztt on
    MaxBacon
  • rertezrertez Member UncommonPosts: 141
    edited September 17
    Xiaoki said:
    MaxBacon said:
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    It's a design ofc, but most do not even try to do one MMO with one different AI system, they just go with the classic design the others do. This has been talked about since the day this game was pitched, this took quite a long time until the AI starting to translate from design to implementation.

    Though I meant the question more related to "the sims" like AI, I'll edit that bit.
    The reason why MMOs go with standard AI is the very nature of MMOs: the internet.

    AI has to be calculated server side because if its calculated client side then it can(and will) be easily hacked to pieces.

    Now, with anything thats done server side you cant calculate it once then send it out to everyone, no, you have to calculate it separately for everyone.

    Lets say 10 people are looking at 1 mob. That means the server has to calculate the AI for that 1 mob 10 times. If you have 10 people looking at 10 mobs then the server has to calculate the AI for those 10 mobs 100 times.

    So, it has never been a matter of companies not wanting or being unable to make better AI, it has always been a matter of having to send millions complex AI calculations to hundreds of thousands of people over the internet every second.

    A couple MMOs have tried complex AI for mobs. Namely, Project Titan by Blizzard and Everquest Next by SOE. Now, who can tell me what happened to them?
    While I agree with your point, note that EQN's emergent AI is the technology driving the mobs' combat behavior in Guild Wars 2's Heart of Thorns expansion and its later raid content aswell. According to the devs' statements after EQN got cancelled the AI part occured to be completely doable and the project was dropped due to completely different reasons.

    For those who are interested in how the AI solutions mentioned above work:
    http://www.gdcvault.com/play/1021848/Building-a-Better-Centaur-AI

    Post edited by rertez on
  • RhoklawRhoklaw Ft. Bliss, TXMember EpicPosts: 5,159
    This is going in the right direction. I know CR wanted a huge persistent universe like Eve Online or maybe even bigger like No Man's Sky, but let's be completely honest. After a while, planets begin to blend in and eventually become repetitive at some point. CIG and CR should focus more on AI mechanics like this and worry less about making a huge open universe. I mean, it should be an open sandbox but you really don't need more than 100 systems or even 50 for that matter.

    QUALITY > QUANTITY
    Octagon7711

  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    hfztt said:
    Yup. The point was to not have to hand animate everything (comparable to hand script missions). So they had code that calculated how to animate the character based on the situation, environment, previous story choices and nature of the dialog (again comparable to the mission setup seen here).

    So yeah. Pretty comparable actually.

    Of course they where up against the fact that reading faces is a thing we humans are very good at, so small deviations seem huge to us. We are much less likely to spot "silly faces" in a mission configuration as we do not have a natural frame of reference for that.
    I didn't know that, I think here they just rely on mocap to trigger the facial animation normally, and fully mocap the NPC idles and such (seeing by Miles).

    rertez said:
    While I agree with your point, note that EQN's emergent AI is the technology driving the mobs' combat behavior in Guild Wars 2's Heart of Thorns expansion and its later raid content aswell. According to the devs' statements after EQN got cancelled the AI part occured to be completely doable and the project was dropped due to completely different reasons.

    For those who are interested in how the AI solutions mentioned above work:
    http://www.gdcvault.com/play/1021848/Building-a-Better-Centaur-AI
    GW2's HoT is emergent AI? I Don't think so, not on that tier of what drives the events and the content generation that comes from their AI, that's the scripted chain that runs in a loop, perhaps something more related to combat?
  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    Rhoklaw said:
    This is going in the right direction. I know CR wanted a huge persistent universe like Eve Online or maybe even bigger like No Man's Sky, but let's be completely honest. After a while, planets begin to blend in and eventually become repetitive at some point. CIG and CR should focus more on AI mechanics like this and worry less about making a huge open universe. I mean, it should be an open sandbox but you really don't need more than 100 systems or even 50 for that matter.

    QUALITY > QUANTITY
    I'm for the same ideology, just to make something big and huge is not what will make it any good due to the daunting repetition factor to put content on it.

    The AI is really what will make or break SC; the core of the game lies on it to ensure the dynamic of the game world and the generation of content that is all to run under this system.

    One of the things I most like of the approach is that if you are stranded in Space/Planet and put out one emergency beacon, the whole system needs to translate that call for help into a mission someone (or AI itself) will take, creating important game loops.
    Rhoklaw
  • ArglebargleArglebargle Austin, TXMember RarePosts: 2,062
    edited September 17
    Absolutely want to see better game AI.  Everything I've seen says it is computationally very expensive though.  Given all the other geegaws in the game, will CIG be able to afford to do extensive AI?  If so, it will be a coup for Star Citizen.

    When I was reading up on the Stalker game series, the devs there talked about putting an advanced AI in for all the NPCs, with the same general goals as the player.   Problem came up that the AI would do everything better and faster than the human player, so they had to dumb it down.
    Post edited by Arglebargle on

    If you are holding out for the perfect game, the only game you play will be the waiting one.

  • XiaokiXiaoki White Pigeon, MIMember RarePosts: 2,872
    Absolutely want to see better game AI.  Everything I've seen says it is computationally very expensive though.  Given all the other geegaws in the game, will CIG be able to afford to do extensive AI?  If so, it will be a coup for Star Citizen.

    When I was reading up on the Stalker game series, the devs there talked about putting an advanced AI in for all the NPCs, with the same general goals as the player.   Problem came up that the AI would do everything better and faster than the human player, so they had to dumb it down.
    Pretty much any developer can make advanced AI that can outsmart and overwhelm a human.

    What makes AI so hard in general is making AI that is smart but not too smart, AI that can be beat and be fun.

    When Naughty Dog was making The Last of Us their first attempt at the AI was completely unbeatable. The enemies would seek cover, flank and support each other. They had to dumb it way down.
  • ExcessionExcession NottinghamMember RarePosts: 709
    edited September 17
    MaxBacon said:
    Specific to subsumption though it's already in for 3.0 so what they talk about is way more fleshed out than what was just one idea back then. This obviously by iterations and the first release is the basic version of what the AI meta is.
    You "think" it is in 3.0, since nobody has played/seen 3.0 yet, nobody knows if it is in or not.
    Post edited by Excession on
    MaxBacon

    A creative person is motivated by the desire to achieve, not the desire to beat others.

  • MaxBaconMaxBacon Figueira da FozMember EpicPosts: 4,071
    edited September 17
    Absolutely want to see better game AI.  Everything I've seen says it is computationally very expensive though.  Given all the other geegaws in the game, will CIG be able to afford to do extensive AI?  If so, it will be a coup for Star Citizen.
    At a profit loss yes.

    Means simply it costs more on servers so they have to assume one higher upkeep to be able to deal with both the complex AI and also physics that are heavy on processing.

    MMO's do this stuff very cheap because the cheaper it is the less of the money earnt will go to pay upkeep costs, and when it comes to MMO companies choosing between profit or better game AI/Physics I think we all know what the decision is, 

    I don't think the challenge comes with making it too smart/balanced, it's its cost, Subsumption is a global system for it so they perhaps can optimize that as most as possible to smartly manage big numbers of AI.

    Excession said:
    You "think" it is in 3.0, since nobody has played/seen 3.0 yet, nobody knows if it is in or not.
    It's not what I think, it's:

    Post edited by MaxBacon on
  • waynejr2waynejr2 West Toluca Lake, CAMember EpicPosts: 7,613
    MaxBacon said:
    How about we get to see it working in game and then we can talk about whether "more games should have a system like this"?

    Talk is cheep.
    It's a design ofc, but most do not even try to do one MMO with one different AI system, they just go with the classic design the others do. This has been talked about since the day this game was pitched, this took quite a long time until the AI starting to translate from design to implementation.

    Though I meant the question more related to "the sims" like AI, I'll edit that bit.

    I don't want a pig in the poke.  You should let it be real before you suck it's ____.
    http://www.youhaventlived.com/qblog/2010/QBlog190810A.html  

    Epic Music:   https://www.youtube.com/watch?v=vAigCvelkhQ&list=PLo9FRw1AkDuQLEz7Gvvaz3ideB2NpFtT1

    https://archive.org/details/softwarelibrary_msdos?&sort=-downloads&page=1

    Kyleran:  "Now there's the real trick, learning to accept and enjoy a game for what it offers rather than pass on what might be a great playing experience because it lacks a few features you prefer."

    John Henry Newman: "A man would do nothing if he waited until he could do it so well that no one could find fault."

    FreddyNoNose:  "A good game needs no defense; a bad game has no defense." "Easily digested content is just as easily forgotten."

    LacedOpium: "So the question that begs to be asked is, if you are not interested in the game mechanics that define the MMORPG genre, then why are you playing an MMORPG?"




  • ExcessionExcession NottinghamMember RarePosts: 709
    MaxBacon said:
    Absolutely want to see better game AI.  Everything I've seen says it is computationally very expensive though.  Given all the other geegaws in the game, will CIG be able to afford to do extensive AI?  If so, it will be a coup for Star Citizen.
    At a profit loss yes.

    Means simply it costs more on servers so they have to assume one higher upkeep to be able to deal with both the complex AI and also physics that are heavy on processing.

    MMO's do this stuff very cheap because the cheaper it is the less of the money earnt will go to pay upkeep costs, and when it comes to MMO companies choosing between profit or better game AI/Physics I think we all know what the decision is, 

    I don't think the challenge comes with making it too smart/balanced, it's its cost, Subsumption is a global system for it so they perhaps can optimize that as most as possible to smartly manage big numbers of AI.

    Excession said:
    You "think" it is in 3.0, since nobody has played/seen 3.0 yet, nobody knows if it is in or not.
    It's not what I think, it's:

    As I said, since nobody has played/seen 3.0, nobody really knows if it is in or not.

    You can link all the screenshot's you like of the 3.0 production schedule, nothing in it has ever been wrong.....
    MaxBaconKefo

    A creative person is motivated by the desire to achieve, not the desire to beat others.

Sign In or Register to comment.