[woutersmits]

i like some ppl to help with this projact
campaign generator its posseble
some saying not but i belive that ai is smarter as us

[Bryan See]

I'm interested in doing that.

[Nightmare]

Already a new project? What about finishing the other things you're working on?

[wookieejedi]

A good idea to recruit help is to lay out a set of goals and larger plan, that way others can see what you are contributing and what specific aspects they might be able to add or help with. I have made my own campaign generator (development is still ongoing 2 years later), and from my experience I could give you a list of lessons I have learned, but I suspect you are looking for someone with more active involvement. From my experience, the amount of time a team leader puts in should be as much if not much more time then the other contributors in the team put in. In other words, leading by example can work wonders.   You also propose using some type of 'ai' in your project, so I would suggest stating what experience you have with AI and what skill sets you possess that would allow this project to reach completion. Are you proposing using a lua solution, custom external script packages, modular missions, etc? Overall, projects with specific goals and a road map are much more likely to attract others and reach a final end product.

Also, I agree with Nightmare. Having too many open projects greatly raises the probability that none of them will ever reach completion. One step at a time with small, doable goals is a good matra to follow 

[potterman28wxcv]

I still think that generating an entertaining random book is easier than this. And nobody generated such a book yet, or at least, not that I'm aware of.

There are a lot progress done in AI field involving neural networks and other stuff. But I think we are still far from AIs having a deep understanding of our language. And this is what is needed if you want to do a campaign generator : you need a way to tell the machine what is a good campaign.

The only way you could do that without doing decades of active research, would be to just take the procedural way : create instances of a more "generic" campaign, using procedural generation. But story-wise it would be in my opinion way more work than redacting an Act for Blue Planet, you would need a lot of fine tuning and a lot of possibilities so that the user keeps getting entertained (in terms of story) with each new campaign..

Obviously the first step to generate a random campaign would be to have a random mission generator. Then you could add constraints/bindings to your random mission generator, and make it so that your random campaign generator uses the random mission generator.

But then the stories would not be original. You wouldn't have unexpected outcomes like you have with the great campaigns out there. It would feel generic and it would be very hard to step out from that impression.

[Nightmare]

Also, there are already tons of exciting campaigns out there, so before you start worrying about a random campaigngenerator - what about playing them all? Many of them have cool ideas, new gameplay features and assets in them.

[wookieejedi]

I still think that generating an entertaining random book is easier than this. And nobody generated such a book yet, or at least, not that I'm aware of.

There are a lot progress done in AI field involving neural networks and other stuff. But I think we are still far from AIs having a deep understanding of our language. And this is what is needed if you want to do a campaign generator : you need a way to tell the machine what is a good campaign.

The only way you could do that without doing decades of active research, would be to just take the procedural way : create instances of a more "generic" campaign, using procedural generation. But story-wise it would be in my opinion way more work than redacting an Act for Blue Planet, you would need a lot of fine tuning and a lot of possibilities so that the user keeps getting entertained (in terms of story) with each new campaign..

Obviously the first step to generate a random campaign would be to have a random mission generator. Then you could add constraints/bindings to your random mission generator, and make it so that your random campaign generator uses the random mission generator.

But then the stories would not be original. You wouldn't have unexpected outcomes like you have with the great campaigns out there. It would feel generic and it would be very hard to step out from that impression.

I agree. An instant action type scenario with slightly altered variables each time would be the most probable, best case outcome in my opinion.

Also, there are already tons of exciting campaigns out there, so before you start worrying about a random campaigngenerator - what about playing them all? Many of them have cool ideas, new gameplay features and assets in them.

^This is a good idea.

[Mongoose]

As an alternative, why not a MadLibs campaign generator?  "It has been XX years since Capella.  Go!"

[wookieejedi]

As an alternative, why not a MadLibs campaign generator?  "It has been XX years since Capella.  Go!"

My immature self would use "Admiral <boogers> just jumped into the <poop> star system to the attack the newly formed <farting> pirate cluster."

[Nightmare]

As an alternative, why not a MadLibs campaign generator?  "It has been XX years since Capella.  Go!"

My immature self would use "Admiral <boogers> just jumped into the <poop> star system to the attack the newly formed <farting> pirate cluster."

Such an idea you can flush down the toilet

[Firesteel]

One area of my research (and my lab's in general) is dynamically picking story elements based on game states. I'd be curious to see how well trying to update a campaign structure based on end state of a mission would go.

[Nightmare]

That's interesting. I think about having a mission generator where the next mission is based upon the outcome of the previous mission. Imagine a new story being created from, say, the destruction of the Colossus in High Noon. Usually you would have to repeat the mission, but this would allow the player to break free from such structures and explore alternate realities... but this is literally fantastic and unlikely to happen...

[Firesteel]

That's the gist of it. At the moment I'm trying to get an interrupt driven conversation system off the ground with the (very long term) goal of being able to use it in combination with that type of reactive planning I talked about. I have no idea if FSO's current version of campaign files are flexible enough to handle either of those two things but I have to get a prototype of something beforehand.

The other thing with FS (at least within my understanding of the current capabilities) is that all those possible missions would need to be hand authored and manually connected.

[karajorma]

Unless you generate the campaign file dynamically too. There's not that much to a campaign file so it would be fairly easy.

[Bryan See]

It's like making use of AI, namely Machine Learning.

[karajorma]

It's really not.

[Bryan See]

Unless it is trained with data for the generator to learn. The data in question is the campaign and missions.

[wookieejedi]

Unless it is trained with data for the generator to learn. The data in question is the campaign and missions.

Machine learning is a bit more complex then that. Also, as mentioned, it is not the same as the dynamic/random campaign or mission generation.

Unless you generate the campaign file dynamically too. There's not that much to a campaign file so it would be fairly easy.

In my personal campaign-esque generator I used that method, but ultimately settled on using mission loops with lua scripted mission rewrites. The more recent SDL file structure came in handy for that.     

[Bryan See]

Unless it is trained with data for the generator to learn. The data in question is the campaign and missions.

Machine learning is a bit more complex then that. Also, as mentioned, it is not the same as the dynamic/random campaign or mission generation.

Unless you generate the campaign file dynamically too. There's not that much to a campaign file so it would be fairly easy.

In my personal campaign-esque generator I used that method, but ultimately settled on using mission loops with lua scripted mission rewrites. The more recent SDL file structure came in handy for that.     

It's something to do with the differences between random generation and pseudo-random generation. The strengths of random generations over the latter is that it is more secure, I think.

I tried to create it, but it requires the FSO codebase.

[Bryan See]

I'm sorry for my previous posts, and I may be thinking of using a Python command-line tool to generate random campaign and missions.