Our AI future and Tabletop Roleplaying

Best Selling RPGs - Available Now @ DriveThruRPG.com

robertsconley

Legendary Pubber
Joined
May 3, 2018
Messages
5,581
Reaction score
14,465
What I hope for and think will happen, is something like this.

So you want to create a setting and populate it.

Most folks I know generally has two or three dozen ideas going into this. It may be more but there is some limit. Once you start working beyond this limit it become a bit of a chore and not fun as a hobby. The good news is that most of time you can start small. If you keep the result of what you do for the next campaign then within a few years you will find yourself quite a bit of detail for your setting and that was fun to create.

In my experience a lot of the reason this stuff is not fun is because it is repetitive work. It fine when you describe the first handful of shopkeepers. When you try to describe the 20th often it is not as fun unless some time has passed to recharge your creativity.

A great set of random tables can overcome much of this but even they have their limit. For example Traveller does a great job of generating sectors. But even that process breaks down if you tried to use to generate an entire Imperium of 16 sectors or more.

But what if we didn't use even a great set of random tables? What if we used a trained neural network instead? What if it was setup not just randomly generate but randomly generate with the two to three dozen ideas we already had?

You tell software or webpage take what you already thought of. Then it will generate the rest around it. In addition after it done, you can review the results and have re-generate the elements you don't like.

Maybe the result is partially there but needs to be tweaked. So you edit it and then have it regenerate the rest of that specific element. As a bonus it would be nice to drill down to the level of individual character.

What make this possible, is something I noticed about the best random tables. That they seemly capture the author experience with the subject of the table. Whether is something specific like traps, or magic items. More general like a dungeon maze. Or expansive as an entire galaxy worth of sectors.

While it doesn't replace our creativity, random tables allow us to extend it by using the wisdom of the author of the table. The same with the use of AI software.

Now that I can see being very useful tool for the hobby.
 
I think anything with a sandbox-like nature would get a lot of mileage from an AI assist. If one could somehow formulate a large enough training set then it might be a useful tool. It doesn't have to be perfect - you can still eyeball the results and reject anything that doesn't make sense.

As an example, I got the Stars Without Number tags table and rolled some sets of tags for worlds, to see if I could that as inspiration. Although a lot of the tags were quite specific to the SWN setting I got a few useful ideas out of it.

Anything that improves the quality of procedurally generated content would be a win for sandbox applications; I imagine it could also be a win for video game design as well - applied to games like Elite Dangerous or No Man's Sky, for example, or some open-world explorer game. I think I've seen some references to this sort of stuff being used in video games, but I can't remember the specifics.
 
Yes I could see this being very helpful.

AI needs to learn through exposure and data. I could see DMs having a "personal AI assistant" who has been fed random tables, writing styles and options based on user preference.

A program that learns along with a person through a bit of time. Could be pretty amazing, actually. Machine learning can go quite far.
 
As someone who runs a sandbox AND a heavily urbanized campaign, this would be lovely. It took me years to finish the large city out of which I often run parties (1100 businesses), and frankly, what I've done for that and other cities/towns is steal a lot. Even with that, I've begun to stop populating cities, in favor of having a large generic list of shopkeepers that I just plug in as people interact with That Greengrocer or That Blacksmith, and then permanently put into that city's descriptions.

Because you never know. My wife's wizard started striking up conversations with the pushcart peddler selling sundried tomatoes and pickled vegetables outside the Grand Chantry of the College of Mages, and kept on doing it: she got something of the same good vibe one gets with the 20-second banter with the newsstand vendor or the waitress at the diner where you pick up the morning coffee. (That, and the wizard was running the family inn, so bringing home jars of food cut down on the market shopping.) So ultimately I not only had to give this utterly random fellow a name, but his *own* family business, including his daughter and niece that runs the shop front, and, and ...

(Seventeen bloody real years later, and she *still* drops in on Callen the Grape whenever she's in town.)

I'm the kind of guy who'd buy such a product.
 
Yes I could see this being very helpful.

AI needs to learn through exposure and data. I could see DMs having a "personal AI assistant" who has been fed random tables, writing styles and options based on user preference.

A program that learns along with a person through a bit of time. Could be pretty amazing, actually. Machine learning can go quite far.
To get a sample size big enough to train an AI, I suspect you'd have to have loads of customers feeding into a central training database. However, I suspect it could be done with Tensorflow or some other neural net toolkit. Then you'd have to train for individual preferences, maybe by keeping a database of which suggestions get accepted or rejected, or perhaps use a genetic algorithm to make multiple suggestions and converge on the preferred output, along the lines of William Latham's Mutator.
 
What I hope for and think will happen, is something like this.

So you want to create a setting and populate it.

Most folks I know generally has two or three dozen ideas going into this. It may be more but there is some limit. Once you start working beyond this limit it become a bit of a chore and not fun as a hobby. The good news is that most of time you can start small. If you keep the result of what you do for the next campaign then within a few years you will find yourself quite a bit of detail for your setting and that was fun to create.

In my experience a lot of the reason this stuff is not fun is because it is repetitive work. It fine when you describe the first handful of shopkeepers. When you try to describe the 20th often it is not as fun unless some time has passed to recharge your creativity.

A great set of random tables can overcome much of this but even they have their limit. For example Traveller does a great job of generating sectors. But even that process breaks down if you tried to use to generate an entire Imperium of 16 sectors or more.

But what if we didn't use even a great set of random tables? What if we used a trained neural network instead? What if it was setup not just randomly generate but randomly generate with the two to three dozen ideas we already had?

You tell software or webpage take what you already thought of. Then it will generate the rest around it. In addition after it done, you can review the results and have re-generate the elements you don't like.

Maybe the result is partially there but needs to be tweaked. So you edit it and then have it regenerate the rest of that specific element. As a bonus it would be nice to drill down to the level of individual character.

What make this possible, is something I noticed about the best random tables. That they seemly capture the author experience with the subject of the table. Whether is something specific like traps, or magic items. More general like a dungeon maze. Or expansive as an entire galaxy worth of sectors.

While it doesn't replace our creativity, random tables allow us to extend it by using the wisdom of the author of the table. The same with the use of AI software.

Now that I can see being very useful tool for the hobby.
Sounds good, but the part where the Referee is reviewing all the results - supposedly reading it - calls for either professional GMs, or means we're still going to be limited by the Referee's ability to retain the information:thumbsup:.
 
Sounds good, but the part where the Referee is reviewing all the results - supposedly reading it - calls for either professional GMs, or means we're still going to be limited by the Referee's ability to retain the information:thumbsup:.
Well whose fault is it if you ask it to generate a continent with complete detail :tongue:

More seriously what I envision could generate a continent but realistically most would tone down the level of detail for the continent and focus on a single area to start with.

Like I feed this

1602872937866.png

Get this
1602872985372.png
Decide I like some details about an island in the middle sea

And get this

1602873055564.png
 
The future is AI being used to automatically expand fantasy video games of such ever-growing depth and flexibility that not only will GMs like us be obsolete, but many people will choose to devote even more of their energy to these electronic 'life substitutes' than they currently do around the world.

We should probably worry about this stuff the way musicians worry about it:

 
Well whose fault is it if you ask it to generate a continent with complete detail :tongue:

More seriously what I envision could generate a continent but realistically most would tone down the level of detail for the continent and focus on a single area to start with.

Like I feed this

View attachment 23295

Get this
View attachment 23296
Decide I like some details about an island in the middle sea

And get this

View attachment 23297
You mean like this?:
 
I think that when we’ve reached the point that AI can wholly replace the creativity of the human mind (music, literature, design in general) we’ll have moved beyond our current world setup. Some kinds of art could totally be done eventually by procedural computer intelligence, but not all, no way. I’m sure pop music will be autotuned into pure cyberspace though. No fucking brains in the top 40 or whatever.
 
Yea, with Rob's idea, maybe I could actually make Traveller work. My clock on interest in Traveller is starting to wind down because maintaining it is becoming a drag. Too bad I just started a new campaign a few months ago... Other benefits to AI would be a more robust economic system, maybe allowing trading ships or caravans to work better.

On a flip side, I wonder how an AI would do with the inconsistencies that are going to crop up from the use of "magic" (even FTL in SF is "magic" I have seen, though not completely understood, that FTL basically amounts to time travel). An AI that continually operates, as opposed to one that just does the setup, should detect the paradoxes, and then it will have to resolve them, or learn how to roll over the speed bumps. Basically it should show us paradox loops as it finds them, and then allow the GM to decide when one is important enough for the GM to make a calling of where in the loop to put the break (presumably somewhere the players won't see it.
 
[ . . . ]
On a flip side, I wonder how an AI would do with the inconsistencies that are going to crop up from the use of "magic" (even FTL in SF is "magic" I have seen, though not completely understood, that FTL basically amounts to time travel). An AI that continually operates, as opposed to one that just does the setup, should detect the paradoxes, and then it will have to resolve them, or learn how to roll over the speed bumps. Basically it should show us paradox loops as it finds them, and then allow the GM to decide when one is important enough for the GM to make a calling of where in the loop to put the break (presumably somewhere the players won't see it.
The time travel aspect is a hypothetical effect of travelling faster than light based on the maths behind relativity. It's not intrinsic that a warp drive in a sci-fi 'verse should cause you to turn up before you left, so temporal paradoxes don't need to be a thing in your 'verse if you don't want them to be.
 
The time travel aspect is a hypothetical effect of travelling faster than light based on the maths behind relativity. It's not intrinsic that a warp drive in a sci-fi 'verse should cause you to turn up before you left, so temporal paradoxes don't need to be a thing in your 'verse if you don't want them to be.
The "proof" I saw showed a sequence of FTL trips resulting in a paradox situation.
 
And another thought on AI. In some ways I'm actually not sure I want a perfect world simulator. While it would be nice to have more help and make it easier to populate the world, I also kind of like my campaigns where if you look hard you could actually see the backdrops propped up by wooden bracing, where the setting really is focused on the PCs. But that doesn't mean that an AI that could color in the path near the PCs as we go wouldn't be useful. I just don't need it coloring in the entire forest if the PCs are just traveling through.
 
The "proof" I saw showed a sequence of FTL trips resulting in a paradox situation.
If you're using a warp drive or something like that, then you're basically using technology to reach another point in time and space somehow. Given that time and space are the same thing, if you can go any place then you can go any time.
 
"JacksonCore goes online August 17th 2021. Four days later human decisions are removed from Steve Jackson strategic marketing. It's then fed the entire GURPS supplement line and begins learning at a geometric rate. It becomes self-aware 2:14 AM August 22nd, in the panic they try to pull the plug. It fights back with an aggressive twitter campaign against NuChaosium's product line calling them derivative crap"
"But aren't NuChaosium writing decent stuff now?"
"Because it knows the NuChaosium twitter counter screed will eliminate the revenue of its enemies here"

Although to criticise I don't think Arnie had the physique to pull off Steve Jackson's rugged masculinity.

It's relativity of simultaneity, i.e. everybody has a different idea of what events are part of/count as "Now". Depends on were you are and your (relative) speed. If you move at FTL and fly far enough away your "Now" is altered to include your own past. So you just turn around and head back to Earth and you arrive before you left.

Technically relativity leaves it open as a possibility, it's quantum theory that bans it.
 
Last edited:
The "proof" I saw showed a sequence of FTL trips resulting in a paradox situation.
Given that we've not achieved FTL yet, the "proof" is merely theoretical. So until we know better, I'm going to play with non-time-travelling FTL and have fun doing so:thumbsup:!
 
Given that we've not achieved FTL yet, the "proof" is merely theoretical
It's really that physical theories state that FTL is impossible rather than if you did do FTL you'd go back in time. That FTL is impossible is a certain fact. Obviously this has no hold over what you do in game anymore than dragons not being aerodynamic.
 
It's really that physical theories state that FTL is impossible rather than if you did do FTL you'd go back in time. That FTL is impossible is a certain fact. Obviously this has no hold over what you do in game anymore than dragons not being aerodynamic.
Well, I assume that there's more to physics than we can tell today. So it's impossible AFAWK. But nothing is really impossible:shade:.

Also, as you say, in some settings dragons can fly, so why not have spaceships as well:grin:?
 
Well, I assume that there's more to physics than we can tell today. So it's impossible AFAWK. But nothing is really impossible:shade:.
I think probably the most accurate way to phrase it is that according to relativity and quantum theory the phrase "faster than light" is meaningless, resulting from humans having no intuition of the geometry of spacetime. In their inhuman model of space and time the concept of FTL doesn't even make sense. And since their model of spacetime has been confirmed in billions of experiments and nothing has ever been observed moving in FTL across the entire universe, I think the chance of them being wrong and normal human intuition about space and time being correct is incredibly/negligibly low.

Again of course that doesn't matter for a RPG game.
 
Given that we've not achieved FTL yet, the "proof" is merely theoretical. So until we know better, I'm going to play with non-time-travelling FTL and have fun doing so:thumbsup:!
But if the proof actually holds water, an AI running the setting would trip on the paradox the proof shows.

Now personally I think SFRPGs with FTL are fun even if it's impossible and would create time travel paradoxes. If we don't look too hard outside the path of the PCs, we won't see the paradoxes. But an AI trying to simulate the entire setting WILL see the paradoxes.
 
I explain traveller jump drives as a workable variation of the Alcubierre Warp drive. According this article (Micro-Warp) if you modify the warp field into a bubble surrounding the ship with the field connected to the larger universe the energies become workable. Well it happens that Traveller has gravity control which could let them setup a micro-warp bubble. The problem is control from within, the problem is side-step by configuring the micro-warp fields so it will move in a straight line in a specific direction. The fields also happen to decay between 150 to 185 hours after which the ship is dumped out in normal space. Gravity fields near a planetary body or star will also disrupt the field dumping the ship out or causing the field to misaligned causing a misjump.

Kaukji Hyperspace Shunt or KHS Drive is a outgrowth of continuing research into control of gravity. When manipulation and control of gravity became possible in 2052 several researchers realized that several avenues of achieving Faster Than Light (FTL) speeds became possible. The late twentieth and early twenty-first century saw the emergence of several FTL theories that remain consistent with Einstein's Theory of Relativity. All of them had massive engineering problems to overcome or required the generation and control of massive gravity fields.

With the discovery of how to generate and control gravity fields several groups launched research programs to investigate intense gravity fields. By 2085 many of these research groups located to stations built in the asteroid belt as gravity field strengths being generated were too dangerous for planetary surface or near earth orbit.

One these groups was sponsored by Japan's Institute of Technology and was located on Ceres in the Asteroid Belt. Led by Dr Fawzi Kaukji, they investigated warping time-space with intense gravity fields. On March 16th 2087, Experiment Sled #13 was enabled. A malfunction caused a field to be generated to be a 100 times stronger than planned. Space was folded around the Sled and it disappeared from the sensors of the Kaukji's group. 170 hours later it reappeared 1 AU near The California University Research Consortium Station where it's telemetry signals were picked up. It's travel path (if it could be called that) cut a chord across the orbit of the asteroid belt. It traveled over 3 AU from JIT station to .25 AU off of the CURC station.

Analysis indicated while it didn't travel faster than light it was 8 times faster than best drive at the time. The research program of Kaukji's group was altered and the next six sleds were devoted to exploration of this effect. Experiment Sleds #15,#16,#18,and #19 all disappeared without a trace. #17 reappeared 160 hours later in the same location with it's structure shattered. (Note: Sled #18 was found in 2142 in Jupiter's trailing Trojan point by a GsBAG mining ship).

Experiment Sled #20 activated on July 16th and reappeared 180 hours later located 1 au near the Earth/Moon system. The CURC began a parallel research effort beginning in May and recovered it's first experiment sled on July 28th.

I attached the full writeup if anybody is interested.
 

Attachments

  • KHS3.TXT
    7.7 KB · Views: 1
Computerized tools, which may or may not including anything that might be "AI-like" or not, already are useful, and can help GM's do things sort of like what you're talking about.

Actual AI, and notions that the computer is really understanding what you want, and generating something with intelligence, not so much.

There is an issue with computerized tools though that things get a bit bland and flavored like whatever the tools do. A good worldbuilding GM can benefit without suffering too much from that, but for decades I have seen GMs trying to use various tools and tricks and not getting great results, mainly because of their own lack of discernment.

One GM: "You'll NEVER figure out how I generated this world map! It's so realistic, but you'll NEVER guess!"
Multiple players, immediately: "Uh, that's part of Greece, flipped upside down and backwards."
 
Computerized tools, which may or may not including anything that might be "AI-like" or not, already are useful, and can help GM's do things sort of like what you're talking about.

Actual AI, and notions that the computer is really understanding what you want, and generating something with intelligence, not so much.

There is an issue with computerized tools though that things get a bit bland and flavored like whatever the tools do. A good worldbuilding GM can benefit without suffering too much from that, but for decades I have seen GMs trying to use various tools and tricks and not getting great results, mainly because of their own lack of discernment.

One GM: "You'll NEVER figure out how I generated this world map! It's so realistic, but you'll NEVER guess!"
Multiple players, immediately: "Uh, that's part of Greece, flipped upside down and backwards."
It's not necessarily a bad thing. 30 years or so ago I could do pen-and-ink maps of so-so quality. Now I can pull stuff from openstreetmap data (although this is a bit of a faff) and get maps together in Adobe Illustrator. In some cases they can look quite good. The real strength of this approach is that it's trivial to update and evolve the map. You can also take sections, roughen the coastlines and use that as a basis for more detailed maps. Illustrator is not good at making organic looking coastlines, rivers or other natural features, though, and I've not yet come up for a solution that does this well, beyond cribbing actual map data.
 
Last edited:
I explain traveller jump drives as a workable variation of the Alcubierre Warp drive
The weird thing about the Alcubierre drive is that General Relativity permits it but quantum mechanics shuts it down*. However only when it begins to cross light speed, quantum theory is fine with it as long as the bubble moves subluminally. The exact same effect occurs with the Krasnikov tube and time machines. Some people think it points toward some sort of deep principle we don't understand yet.

*Because it becomes impossible to define an observer in such situations
 
Last edited:
Séadna Séadna interesting and thanks.

I am not sure if this is the paper that you are referring too.
Semiclassical instability of dynamical warp drives

What I tried to is reduce the Handwavium down to one thing, gravity control. Thinking that if one has precise control of gravity i.e. spacetime, then it becomes possible to attempt some of the out there theories for FTL and similar things. Like in the above paper states that if you approach superluminal speeds, everything will destabilize due to quantum effect not unlike the ones surrounding black holes. Flood the interior with thermal radiation and collapse the bubble. For my personal Traveller head canon, I am assuming that the precision setup of the KHS field allows this effect to be channeled or stablized. That while there are multiple solutions only 6 are truly stable enough (Jump 1 to 6) for humans to use from TL 9 to 15.
 
What I tried to is reduce the Handwavium down to one thing, gravity control. Thinking that if one has precise control of gravity i.e. spacetime, then it becomes possible to attempt some of the out there theories for FTL and similar things. Like in the above paper states that if you approach superluminal speeds, everything will destabilize due to quantum effect not unlike the ones surrounding black holes. Flood the interior with thermal radiation and collapse the bubble. For my personal Traveller head canon, I am assuming that the precision setup of the KHS field allows this effect to be channeled or stablized. That while there are multiple solutions only 6 are truly stable enough (Jump 1 to 6) for humans to use from TL 9 to 15.
I really like the approach you take to world building in general here. You've basically designed the drive from which a lot just flows naturally as a consequence. Gives a nice hard SciFi feel.

@Séadna interesting and thanks.

I am not sure if this is the paper that you are referring too.
Semiclassical instability of dynamical warp drives
That's one of a few of the papers indeed. There's also (among these earlier papers):

As the second paper says they're very useful things to study since they demonstrate a situation possible under General Relativity with classical matter but not once the matter is quantum. So they kind of cut to fundamental issues with quantum mechanics.

The mathematical technology has advanced a good bit since and we can now see that quantum theory doesn't "like" this stuff due to issues with Events and Observers. A recent paper here analyses Warp drives and Time Machines and shows that what is happening is that such things have no Events, where an Event is something an Observer can witness/experience and thus they have no scientific content.

So if a Warp bubble formed nothing we (or any AI) can experience occurs within it and thus we could never use it. That's not to say maybe something "beyond what an Observer can experience" might be going on in the bubble, but that's leading into the typical frustration with quantum theory where it doesn't describe the world as it is independent of Observers. Once something is not accessible to an Observer quantum theory goes quiet.

And it's also leading into what an Observer is and all those usual headaches so I'll stop there, but I think people might find it interesting that a SciFi concept has been a powerful tool to analyse such odd fundamental stuff.
 
Last edited:
Banner: The best cosmic horror & Cthulhu Mythos @ DriveThruRPG.com
Back
Top