Jump to content

Recommended Posts

Posted

Has anyone tried to create missions using AI such as OpenAI etc? Even more importantly, has anyone successfully created a mission using AI? Is it even possible? I had a go, but it seems to hard to do.

  • Like 1
Posted
12 minutes ago, filthymanc said:

Has anyone tried to create missions using AI such as OpenAI etc?

 

I use ChatGPT to write for me  fictional storylines for some of my missions, it's very good at that. With some effort I can get complete storylines 🙂 

 

12 minutes ago, filthymanc said:

Even more importantly, has anyone successfully created a mission using AI? Is it even possible?

 

I have not and believe that most likely it is not feasible.

  • Like 2

 

For work: iMac mid-2010 of 27" - Core i7 870 - 6 GB DDR3 1333 MHz - ATI HD5670 - SSD 256 GB - HDD 2 TB - macOS High Sierra

For Gaming: 34" Monitor - Ryzen 3600 - 32 GB DDR4 2400 - nVidia RTX2080 - SSD 1.25 TB - HDD 10 TB - Win10 Pro - TM HOTAS Cougar

Mobile: iPad Pro 12.9" of 256 GB

Posted
5 minutes ago, Rudel_chw said:

I use ChatGPT to write for me  fictional storylines for some of my missions, it's very good at that. With some effort I can get complete storylines 🙂 

Agreed, it does have a better imagination than me.

  • Like 2
Posted
5 minutes ago, filthymanc said:

Agreed, it does have a better imagination than me.

 

Me too 🙂 ... here is a sample:

 

VuyRA97.jpg

 

  • Like 3

 

For work: iMac mid-2010 of 27" - Core i7 870 - 6 GB DDR3 1333 MHz - ATI HD5670 - SSD 256 GB - HDD 2 TB - macOS High Sierra

For Gaming: 34" Monitor - Ryzen 3600 - 32 GB DDR4 2400 - nVidia RTX2080 - SSD 1.25 TB - HDD 10 TB - Win10 Pro - TM HOTAS Cougar

Mobile: iPad Pro 12.9" of 256 GB

Posted
5 hours ago, filthymanc said:

Has anyone tried to create missions using AI such as OpenAI etc? Even more importantly, has anyone successfully created a mission using AI? Is it even possible? I had a go, but it seems to hard to do.

I think it may work well for creating a storyline, but as for actually creating the mission good luck.  Current AI are still too prone to making stuff up if it doesn't know the answer so most of the code it comes up with is pretty questionable.

Posted

It's pretty simple XML though isn't it?

mission = 
 

Spoiler

{
    ["groundControl"] = 
    {
        ["passwords"] = 
        {
            ["artillery_commander"] = {},
            ["instructor"] = {},
            ["observer"] = {},
            ["forward_observer"] = {},
        }, -- end of ["passwords"]
        ["roles"] = 
        {
            ["artillery_commander"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["artillery_commander"]
            ["instructor"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["instructor"]
            ["observer"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["observer"]
            ["forward_observer"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["forward_observer"]
        }, -- end of ["roles"]
        ["isPilotControlVehicles"] = false,
    }, -- end of ["groundControl"]
    ["requiredModules"] = {},
    ["date"] = 
    {
        ["Day"] = 12,
        ["Year"] = 1982,
        ["Month"] = 1,
    }, -- end of ["date"]
    ["trig"] = 
    {
        ["actions"] = 
        {
            [1] = "a_effect_smoke(10, 1, 0.05, \"\");a_effect_smoke(20, 1, 0.05, \"\"); mission.trig.func[1]=nil;",
        }, -- end of ["actions"]
        ["events"] = {},
        ["custom"] = {},
        ["func"] = 
        {
            [1] = "if mission.trig.conditions[1]() then mission.trig.actions[1]() end",
        }, -- end of ["func"]
        ["flag"] = 
        {
            [1] = true,
        }, -- end of ["flag"]
        ["conditions"] = 
        {
            [1] = "return(c_time_before(100) )",
        }, -- end of ["conditions"]
        ["customStartup"] = {},
        ["funcStartup"] = {},
    }, -- end of ["trig"]
    ["maxDictId"] = 5,
    ["result"] = 
    {
        ["offline"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["offline"]
        ["total"] = 0,
        ["blue"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["blue"]
        ["red"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["red"]
    }, -- end of ["result"]
    ["pictureFileNameN"] = {},
    ["drawings"] = 
    {
        ["options"] = 
        {
            ["hiddenOnF10Map"] = 
            {
                ["Observer"] = 
                {
                    ["Neutral"] = false,

 

Cheers.

Posted
1 hour ago, Weta43 said:

It's pretty simple XML though isn't it?

mission = 
 

  Reveal hidden contents

{
    ["groundControl"] = 
    {
        ["passwords"] = 
        {
            ["artillery_commander"] = {},
            ["instructor"] = {},
            ["observer"] = {},
            ["forward_observer"] = {},
        }, -- end of ["passwords"]
        ["roles"] = 
        {
            ["artillery_commander"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["artillery_commander"]
            ["instructor"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["instructor"]
            ["observer"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["observer"]
            ["forward_observer"] = 
            {
                ["neutrals"] = 0,
                ["blue"] = 0,
                ["red"] = 0,
            }, -- end of ["forward_observer"]
        }, -- end of ["roles"]
        ["isPilotControlVehicles"] = false,
    }, -- end of ["groundControl"]
    ["requiredModules"] = {},
    ["date"] = 
    {
        ["Day"] = 12,
        ["Year"] = 1982,
        ["Month"] = 1,
    }, -- end of ["date"]
    ["trig"] = 
    {
        ["actions"] = 
        {
            [1] = "a_effect_smoke(10, 1, 0.05, \"\");a_effect_smoke(20, 1, 0.05, \"\"); mission.trig.func[1]=nil;",
        }, -- end of ["actions"]
        ["events"] = {},
        ["custom"] = {},
        ["func"] = 
        {
            [1] = "if mission.trig.conditions[1]() then mission.trig.actions[1]() end",
        }, -- end of ["func"]
        ["flag"] = 
        {
            [1] = true,
        }, -- end of ["flag"]
        ["conditions"] = 
        {
            [1] = "return(c_time_before(100) )",
        }, -- end of ["conditions"]
        ["customStartup"] = {},
        ["funcStartup"] = {},
    }, -- end of ["trig"]
    ["maxDictId"] = 5,
    ["result"] = 
    {
        ["offline"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["offline"]
        ["total"] = 0,
        ["blue"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["blue"]
        ["red"] = 
        {
            ["conditions"] = {},
            ["actions"] = {},
            ["func"] = {},
        }, -- end of ["red"]
    }, -- end of ["result"]
    ["pictureFileNameN"] = {},
    ["drawings"] = 
    {
        ["options"] = 
        {
            ["hiddenOnF10Map"] = 
            {
                ["Observer"] = 
                {
                    ["Neutral"] = false,

 

Assuming the AI actually understood that, yes.  Problem is that currently most of them literally just make stuff up if they aren't sure and it ends up being gibberish.  At least that's what I've seen reported so far.

  • Like 1
Posted

It's very specific XML with a fixed syntax, and there's not enough material to train a dedicated AI for it. It's exactly what AI tends to be very bad at.

I'd be careful with that story, too. It's too short to make a call on the actual story (other than the writing is excessively verbose), but I'm pretty sure you'd want a callsign that makes sense.

  • Like 1
  • 2 months later...
Posted

I use it with the MOOSE Framework, with positive results. Granted the missions are somewhat simple in that they are merely solo practice, and don't carry much complexity, but the potential is there. As my scripting powers accumulate, I will be more adept at refining the tool in the future so I believe it only gets better.

ChatGPT whipped up a simple JTAC practice mission that worked "out of the box" this morning.

To circle back to the question, my answer is 'Yes' you can use OpenAI to create working missions.

 

 

 

 

  • Like 2
Posted

I suppose if you keep it simple it might work. Quite frankly, I'm surprised it didn't butcher it completely. It's definitely interesting, I wonder if it can make one using realistic procedures (time on target, push times and so on). It probably has a limit, but it could save a lot of tedious work in ME. 

I still wouldn't use it for writing stories, though. Unless, of course, you want the most droll, cookie cutter story ridden with purple prose, hackneyed tropes, bad "research" and overall poor writing that you ever saw outside sitcoms and mass-produced crime novels. 🙂 

Posted

My current practice mission is evolving everyday, yesterday was adding the keyhole positions as I just learned about what that is and how to navigate one, awesome stuff.

I will have to dig into what the scripting side will actually support for full JTAC, which would be amazing. My current testing is around the IP points and how JTAC will see them and react with them in game. Surprisingly very little documentation on IP points, had to look it up in the old A10 manual, lol.

As for story writing, 100% agree, might as well start every story with 'Once upon a time, there was an F18...eww' 😉

I think we can train it to produce better results, but full literary chops, probably not for a long time.

Posted

I have used ai to successfully mission briefings and create starting locations, ground frag orders, and air sorties based on last mission completion state. It is generally working as a dynamic campaign generator by acting as the red force and blue force commanders. Unfortunately it still requires tons of manual work to build each mission for all the ground movement and air sorties with the actual mission editor. It basically just says what the groups will do and where they will attack and how. 

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...