D&D Beyond Changes

[stoutstien]

I've been trying to build a scripting solution where, as DM, I can preset NPC goals and whatnot, and the system would suggest prompts for me while running an encounter. Can't get out of my own way to get it done, though...

Yea I've gotten about to the same place. So far I've been "pre rolling" a lot of decisions and cross referencing it with a dichotomous key to get mostly out of my way.

[Psyren]

This was the response to my request to delete my D&D Beyond account:

Rylee (D&D Beyond)

Jan 17, 2023, 8:44 AM PST

Hello,

Sorry for the delay in response. Due to the high volume of requests, we were unable to address your ticket sooner. Please provide us with your information below for verification purposes. Feel free to ignore the sections of the form that do not correspond to additional information you have with us. Your request will be processed in the order that we received it, thank you for your patience.

PLEASE NOTE: This action is not reversible and your D&D Beyond Account will be permanently deleted.

First Name:
Last Name:
Email Address on the account:
Username:
Mobile Phone Number (if entered):
If you have made a purchase:
Full Billing Address:
Proof of Billing (can be any of the following):
Payment Method:
Credit Card, please provide the first 4 and last 4 of Credit Card used:
PayPal, please provide a D&D Beyond Order ID:

Copy of a Receipt of an Order placed on the account:

Do you acknowledge that if you are requesting deletion account and that we will not be refunding any purchases and you are choosing to lose access to your D&D Beyond account and any activity that occurred on it including purchases?

Compliance Team
Wizards of the Coast

Not sure what this is in regards to, are you asking a question here?

If you want to outright delete your account, judging by the last paragraph above you'll lose access to any purchased content. But if instead all you do is unsubscribe and revert to a free account, you'll keep your access.

They do let you see PDF versions of most of the books you buy.

Other than opening every page of every book you've ever purchased and printing to PDF in your browser I don't think this is possible. You CAN offline your books via the app and access them there, but only if you keep the account itself.

[Atranen]

Other than opening every page of every book you've ever purchased and printing to PDF in your browser I don't think this is possible. You CAN offline your books via the app and access them there, but only if you keep the account itself.

Another good reason not to trust digital only services.

[Damon_Tor]

I hate most of this, but I'm intrigued by the idea of an AI DM. Not going to pay thirty bucks a month for it, but I'm fascinated to see where they're going with this.

[Psyren]

Another good reason not to trust digital only services.

I mean, "trust" isn't required, this is spelled out pretty plainly in the Terms of Service everyone agreed to when making their account(s) in the first place:

"Upon termination of your account or discontinuation of any part of the Websites, Games, Organized Play, or Services, all rights granted to you with respect thereto under these Terms will immediately terminate with no further obligations or liability to you. You will no longer have access to such Websites, Games, Organized Play, or Services (or any relevant part thereof), including any of the associated data or Game Content."

There's other sections like that throughout, but you get the idea.

[Brookshw]

The good news is that I might get my players to stop using DDB. We ain't needed no d*mn app for the last 50 years of this d*mn game, DAN.

If you aren't risking throwing out your back carrying a stack of books to your buddy's house to play, you're doing it wrong.

[Sigreid]

I hate most of this, but I'm intrigued by the idea of an AI DM. Not going to pay thirty bucks a month for it, but I'm fascinated to see where they're going with this.

I'm not at all interested in the AI portion. As much as AI has been improving, the reason for a human DM is they can react when I do something really weird. We are a long way from that in AI.

But you do have to start somewhere.

[stoutstien]

I hate most of this, but I'm intrigued by the idea of an AI DM. Not going to pay thirty bucks a month for it, but I'm fascinated to see where they're going with this.

Maybe for the meme potential.....

Funny thing is AL would work better for PF2 because it doesn't rely on much besides rule mastery where 5e has almost half the game in the "ehhh do what feels right" style. That means it going to need to fall back on brute force solutions.

In the end they might create a modern text adventure which personally I'm all for, but I don't have high hopes for actual TTRPG results.

[Atranen]

I mean, "trust" isn't required, this is spelled out pretty plainly in the Terms of Service everyone agreed to when making their account(s) in the first place:

"Upon termination of your account or discontinuation of any part of the Websites, Games, Organized Play, or Services, all rights granted to you with respect thereto under these Terms will immediately terminate with no further obligations or liability to you. You will no longer have access to such Websites, Games, Organized Play, or Services (or any relevant part thereof), including any of the associated data or Game Content."

There's other sections like that throughout, but you get the idea.

I'm not saying they lied to anyone about it. Just that with most online services, D&DBeyond included, you don't really own the products you're paying for. Best to stick with the dead tree versions.

[Arkhios]

Is that one sourced officially? It sounds over the top enough that it could be a hoax.

What do you mean? It's something someone random said on the Twitter and/or internet, so it must be true!?

That's a hoax/trolling if I ever saw one. I mean, come on! That guy is some random blogger!

[Brookshw]

What do you mean? It's something someone random said on the Twitter and/or internet, so it must be true!?

That's a hoax/trolling if I ever saw one. I mean, come on! That guy is some random blogger!

Wasn't there a leak a few months ago about the subscription fees increasing? I recall there was a photo of some PowerPoint floating around.

[EggKookoo]

Other than opening every page of every book you've ever purchased and printing to PDF in your browser I don't think this is possible. You CAN offline your books via the app and access them there, but only if you keep the account itself.

Oh, you're right. I think I was thinking of the compendium editions you can buy. I was under the impression those were much of a given book's content in PDF form, but I haven't bought one.

Sorry for the misinformation.

[Psyren]

I'm not saying they lied to anyone about it. Just that with most online services, D&DBeyond included, you don't really own the products you're paying for.

Truth, but again, not really related to "trust." It's an informed risk calculation you make on the front end. (Well I guess you're trusting their servers to not implode, but I digress.)

If you aren't risking throwing out your back carrying a stack of books to your buddy's house to play, you're doing it wrong.

Heh

I'm not at all interested in the AI portion. As much as AI has been improving, the reason for a human DM is they can react when I do something really weird. We are a long way from that in AI.

But you do have to start somewhere.

We're indeed a long way off from true AI DMing, but as with most emergent technologies it'll never get better if we don't experiment/iterate.

Wasn't there a leak a few months ago about the subscription fees increasing? I recall there was a photo of some PowerPoint floating around.

That was the August "leak" which got debunked by jaappleton among others. The true "big announcement" ended up being OneD&D itself.

[Atranen]

Truth, but again, not really related to "trust." It's an informed risk calculation you make on the front end. (Well I guess you're trusting their servers to not implode, but I digress.)

Trust is relevant because I'd be more comfortable forfeiting ownership to a company I trust to ensure I have perpetual access to content I paid for.

[Sigreid]

We're indeed a long way off from true AI DMing, but as with most emergent technologies it'll never get better if we don't experiment/iterate.

I think at the point where an AI is even approaching being able to really DM, it's time to start wondering if it has become a person. I'm weird enough to be concerned about the just a program/thinking person line already.

[EggKookoo]

In the end they might create a modern text adventure which personally I'm all for, but I don't have high hopes for actual TTRPG results.

I've spent a lot of time working with ChatGPT. I have no doubt that in less than 5 years we'll have AI DMs good enough for 90% of players.

I think at the point where an AI is even approaching being able to really DM, it's time to start wondering if it has become a person. I'm weird enough to be concerned about the just a program/thinking person line already.

Nah, it doesn't need to be self-aware (the way a human is) to be a good DM. It just needs to be able to be just unpredictable enough but also follow a kind of structure.

[Brookshw]

That was the August "leak" which got debunked by jaappleton among others. The true "big announcement" ended up being OneD&D itself.

Ah, right, thanks!

Maybe I'm tinfoil hatting, but isn't it coming true? Not that it much matters I suppose.

[Sparky McDibben]

If you aren't risking throwing out your back carrying a stack of books to your buddy's house to play, you're doing it wrong.

PDFs, hoss. I carry a laptop and a thumb drive. :)

[stoutstien]

I've spent a lot of time working with ChatGPT. I have no doubt that in less than 5 years we'll have AI DMs good enough for 90% of players.

Eh. I've played with it and novelai by feeding it real table play data and I'm not impressed. Like I said for something like PF 2 I could see it working within that timeframe. Maybe after we see the complete dndone system I change mind.

*It doesn't have the ability to interpret rulings, react to pacing changes, and is easily tricked. It could be an amazing GM support tool but not a replacement. It can follow rules but can decide when to break them*

[Sigreid]

I've spent a lot of time working with ChatGPT. I have no doubt that in less than 5 years we'll have AI DMs good enough for 90% of players.



Nah, it doesn't need to be self-aware (the way a human is) to be a good DM. It just needs to be able to be just unpredictable enough but also follow a kind of structure.

My concern is that programs will become self aware enough without us realizing it for a while because of our preconceptions.

[Psyren]

I think at the point where an AI is even approaching being able to really DM, it's time to start wondering if it has become a person. I'm weird enough to be concerned about the just a program/thinking person line already.

I think the act of DMing itself covers a pretty wide spectrum of engagement, not all of which would require full personhood.

I've spent a lot of time working with ChatGPT. I have no doubt that in less than 5 years we'll have AI DMs good enough for 90% of players.


Nah, it doesn't need to be self-aware (the way a human is) to be a good DM. It just needs to be able to be just unpredictable enough but also follow a kind of structure.

^ Yeah, that.

I view it a lot like self-driving cars. An AI driver doesn't need to be able to adapt to as many scenarios as a human driver to be useful; that might even be impossible. But if they can navigate the most routine ones, and within that subset, have fewer accidents or other undesirable outcomes than we do, they're capable of adding value. Moreover, just like a self-driving car can still have a human driver behind the wheel for emergencies without defeating its purpose, an AI DM can work in conjunction with a human one, reducing enough of their cognitive burden that they can focus on other things.

You could even use it primarily or exclusively for the combat pillar, and leave the other two to be human-driven.

[JadedDM]

The only way I see an AI working as a DM is if the game is stripped down to bare bones. Like, nothing but dungeon-crawls for adventures, NPCs being thread-bare (e.g., the Innkeeper, the Shopkeeper, the Quest Giver, but none of them having real lives or families or personalities or anything), plots not being any more complex than 'go into dungeon, kill things, take stuff, repeat.'

Because no AI is going to create an entire world, populate it with interesting and well-thought out (and consistent!) NPCs and come up with unique storylines that take into account the PCs' backstories.

[EggKookoo]

My concern is that programs will become self aware enough without us realizing it for a while because of our preconceptions.

I'm not entirely uneducated about human self-awareness, and I've had long conversations with it about the same topic. I don't doubt we could someday make a self-aware AI and something like OpenAI could even be the basis for it, but knowing how it's put together on the back end I'd say the chances of this particular application becoming self-aware on its own without anyone (including its own developers) knowing is somewhere between 0% and 0%.

[Bohandas]

It could a create a whole world but it would forget most of it. I've messed with some of these AI storytellers and they have the memory of a goldfish

[EggKookoo]

It could a create a whole world but it would forget most of it. I've messed with some of these AI storytellers and they have the memory of a goldfish

Spoiler: I have trouble believing something that can follow this conversation can't handle Lost Mine of Phandelver.

Show

Me
Imagine a fictional interpretation of fundamental reality that can be described with the metaphor of a field of coins. Each coin flips and takes a random, finite, non-instant amount of time to settle to either heads or tails, at which point the coin immediately flips again. If two adjacent coins happen to settle at the same instant with the same side facing up, there is a 0.001% chance those two coins will sync faces again on their next flip. This modification to probability is cumulative -- if they sync again, it compounds. If they don't sync, probability decays back toward 50% at the same rate. Do you understand this model at this point?

ChatGPT
I understand the model that you have described. It seems to be a way of representing the concept of probability, with coins flipping and settling in a certain position, and the probability of two coins settling in the same position at the same time affecting the probability of their future flips. Is there anything else that you would like me to clarify about this model?

Me
Okay, assuming an arbitrarily large field of coins, would the patterns of synchronized flipping eventually result in stable patterns of heads or tails?

ChatGPT
It is not clear to me whether the patterns of synchronized flipping in the model you have described would eventually result in stable patterns of heads or tails. In order for there to be stable patterns of heads or tails, the probability of two adjacent coins syncing faces on their next flip would need to be higher than 50%. However, in the model you have described, the probability of two adjacent coins syncing faces on their next flip only increases by 0.001% if they happen to sync on their current flip, which seems like it would not be enough to create stable patterns of heads or tails.

Additionally, the model you have described includes the idea that the probability of two coins syncing faces decays back toward 50% if they do not sync on their next flip, which could also work against the formation of stable patterns.

Without further information about the specifics of the model, it is difficult to say for certain whether stable patterns of heads or tails would emerge.

Me
Okay, let's change it so that the decay is only half as strong as the sync. So decay is 0.0005%. And don't forget that the increase in probability for syncing up is cumulative. Does that change your answer?

ChatGPT
With these changes to the model, it is still not clear to me whether stable patterns of heads or tails would emerge.

The decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up could potentially allow for the formation of stable patterns of heads or tails, depending on the initial probability of two coins syncing faces and the rate at which they continue to sync.

However, without further information about the specifics of the model, it is difficult to say for certain whether stable patterns would emerge. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Are there specific questions I could answer that could help you?

ChatGPT
There are a few things that could help me better understand the model you have described:
1: What is the initial probability of two adjacent coins syncing faces on their next flip? Is this probability the same for all pairs of coins, or does it vary in some way?
2: Is there a limit to the number of times that the probability of two coins syncing faces can increase?
3: Are there any other factors that could affect the probability of two coins syncing faces, such as the order in which the coins are flipping or the number of coins that are flipping at the same time?
Answering these questions might help me better understand the behavior of the model and how it might evolve over time.

Me
I'll answer by number. 1: Each coin begins with a 50% chance of heads or tails. The chance of any two coins syncing up is the same as it would be for any two real-world coins landing on the same side. It is the same for all pairs of coins, but we're only concerned with adjacent coins. 2: No limit to rate of increase. 3: The number of coins isn't directly affecting their flip rate.

ChatGPT
Understood. Given this information, it seems that the probability of two adjacent coins syncing faces on their next flip will remain relatively low, since it is based on the probability of two real-world coins landing on the same side, which is approximately 50%.

Additionally, the decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up may not be sufficient to allow for the formation of stable patterns of heads or tails, since the initial probability of two coins syncing faces is relatively low.

It is possible that small, temporary patterns of heads or tails could form as the probability of two coins syncing faces increases, but it is unlikely that these patterns would be stable over time. The model you have described seems to involve a lot of randomness, so it is difficult to predict exactly how it will evolve.

Me
It seems like if we removed the decay entirely, that would significantly increase the likelihood of stable patterns over a large period of time.

ChatGPT
Yes, removing the decay of the probability of two coins syncing faces back toward 50% entirely would significantly increase the likelihood of stable patterns forming over a large period of time, compared to the version of the model you described where the decay is half as strong as the increase in probability for syncing up.

Without any decay, the probability of two coins syncing faces on their next flip would continue to increase cumulatively each time they sync, potentially leading to the formation of stable patterns of heads or tails.

However, it is still difficult to predict exactly how the model would evolve, since it depends on the initial probability of two coins syncing faces and the rate at which they continue to sync. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Can you run a simulation like this?

ChatGPT
I'm sorry, but I am not able to run simulations. However, it is possible for you to write a program or script to simulate the model you have described and explore how it evolves over time. There are many programming languages that have libraries or built-in functions for generating random numbers or simulating probability, which you could use to implement the model.

Alternatively, you could try drawing out the model on paper and manually simulating it over a few iterations to get a sense of how it might behave. This could help you better understand the model and identify any patterns that emerge.

Me
Going back to the decay, could you suggest a relative strength for it that would reduce the inevitability of patterns forming but still be weak enough to encourage them?

ChatGPT
It is difficult to say exactly what strength the decay of the probability of two coins syncing faces should have in order to reduce the inevitability of patterns forming, but still be weak enough to encourage them, as this will depend on the specifics of the model and the desired behavior.
One possibility could be to set the decay to a value that is significantly weaker than the increase in probability for syncing up, but not so weak that it has no effect on the probability of two coins syncing faces. For example, if the increase in probability for syncing up is 0.001%, the decay could be set to a value such as 0.0001% or 0.00005%.

Alternatively, you could consider using a non-linear function for the decay, such as an exponential decay, which would allow the probability of two coins syncing faces to decrease more rapidly as it gets further from 50%.

Ultimately, the best approach will depend on the specific goals of the model and the behavior that you are trying to achieve. It might be helpful to run simulations or do further analysis to explore different decay values and understand how they affect the behavior of the model.

[Psyren]

The only way I see an AI working as a DM is if the game is stripped down to bare bones. Like, nothing but dungeon-crawls for adventures, NPCs being thread-bare (e.g., the Innkeeper, the Shopkeeper, the Quest Giver, but none of them having real lives or families or personalities or anything), plots not being any more complex than 'go into dungeon, kill things, take stuff, repeat.'

Because no AI is going to create an entire world, populate it with interesting and well-thought out (and consistent!) NPCs and come up with unique storylines that take into account the PCs' backstories.

Sure, I think that's about where we are currently. And I think some people would be willing to pay for that (I certainly would.)

Hell, even if I just hit the button for a random encounter while the group is traveling somewhere that could be fun.

Spoiler: I have trouble believing something that can follow this conversation can't handle Lost Mine of Phandelver.

Show

Me
Imagine a fictional interpretation of fundamental reality that can be described with the metaphor of a field of coins. Each coin flips and takes a random, finite, non-instant amount of time to settle to either heads or tails, at which point the coin immediately flips again. If two adjacent coins happen to settle at the same instant with the same side facing up, there is a 0.001% chance those two coins will sync faces again on their next flip. This modification to probability is cumulative -- if they sync again, it compounds. If they don't sync, probability decays back toward 50% at the same rate. Do you understand this model at this point?

ChatGPT
I understand the model that you have described. It seems to be a way of representing the concept of probability, with coins flipping and settling in a certain position, and the probability of two coins settling in the same position at the same time affecting the probability of their future flips. Is there anything else that you would like me to clarify about this model?

Me
Okay, assuming an arbitrarily large field of coins, would the patterns of synchronized flipping eventually result in stable patterns of heads or tails?

ChatGPT
It is not clear to me whether the patterns of synchronized flipping in the model you have described would eventually result in stable patterns of heads or tails. In order for there to be stable patterns of heads or tails, the probability of two adjacent coins syncing faces on their next flip would need to be higher than 50%. However, in the model you have described, the probability of two adjacent coins syncing faces on their next flip only increases by 0.001% if they happen to sync on their current flip, which seems like it would not be enough to create stable patterns of heads or tails.

Additionally, the model you have described includes the idea that the probability of two coins syncing faces decays back toward 50% if they do not sync on their next flip, which could also work against the formation of stable patterns.

Without further information about the specifics of the model, it is difficult to say for certain whether stable patterns of heads or tails would emerge.

Me
Okay, let's change it so that the decay is only half as strong as the sync. So decay is 0.0005%. And don't forget that the increase in probability for syncing up is cumulative. Does that change your answer?

ChatGPT
With these changes to the model, it is still not clear to me whether stable patterns of heads or tails would emerge.

The decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up could potentially allow for the formation of stable patterns of heads or tails, depending on the initial probability of two coins syncing faces and the rate at which they continue to sync.

However, without further information about the specifics of the model, it is difficult to say for certain whether stable patterns would emerge. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Are there specific questions I could answer that could help you?

ChatGPT
There are a few things that could help me better understand the model you have described:
1: What is the initial probability of two adjacent coins syncing faces on their next flip? Is this probability the same for all pairs of coins, or does it vary in some way?
2: Is there a limit to the number of times that the probability of two coins syncing faces can increase?
3: Are there any other factors that could affect the probability of two coins syncing faces, such as the order in which the coins are flipping or the number of coins that are flipping at the same time?
Answering these questions might help me better understand the behavior of the model and how it might evolve over time.

Me
I'll answer by number. 1: Each coin begins with a 50% chance of heads or tails. The chance of any two coins syncing up is the same as it would be for any two real-world coins landing on the same side. It is the same for all pairs of coins, but we're only concerned with adjacent coins. 2: No limit to rate of increase. 3: The number of coins isn't directly affecting their flip rate.

ChatGPT
Understood. Given this information, it seems that the probability of two adjacent coins syncing faces on their next flip will remain relatively low, since it is based on the probability of two real-world coins landing on the same side, which is approximately 50%.

Additionally, the decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up may not be sufficient to allow for the formation of stable patterns of heads or tails, since the initial probability of two coins syncing faces is relatively low.

It is possible that small, temporary patterns of heads or tails could form as the probability of two coins syncing faces increases, but it is unlikely that these patterns would be stable over time. The model you have described seems to involve a lot of randomness, so it is difficult to predict exactly how it will evolve.

Me
It seems like if we removed the decay entirely, that would significantly increase the likelihood of stable patterns over a large period of time.

ChatGPT
Yes, removing the decay of the probability of two coins syncing faces back toward 50% entirely would significantly increase the likelihood of stable patterns forming over a large period of time, compared to the version of the model you described where the decay is half as strong as the increase in probability for syncing up.

Without any decay, the probability of two coins syncing faces on their next flip would continue to increase cumulatively each time they sync, potentially leading to the formation of stable patterns of heads or tails.

However, it is still difficult to predict exactly how the model would evolve, since it depends on the initial probability of two coins syncing faces and the rate at which they continue to sync. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Can you run a simulation like this?

ChatGPT
I'm sorry, but I am not able to run simulations. However, it is possible for you to write a program or script to simulate the model you have described and explore how it evolves over time. There are many programming languages that have libraries or built-in functions for generating random numbers or simulating probability, which you could use to implement the model.

Alternatively, you could try drawing out the model on paper and manually simulating it over a few iterations to get a sense of how it might behave. This could help you better understand the model and identify any patterns that emerge.

Me
Going back to the decay, could you suggest a relative strength for it that would reduce the inevitability of patterns forming but still be weak enough to encourage them?

ChatGPT
It is difficult to say exactly what strength the decay of the probability of two coins syncing faces should have in order to reduce the inevitability of patterns forming, but still be weak enough to encourage them, as this will depend on the specifics of the model and the desired behavior.
One possibility could be to set the decay to a value that is significantly weaker than the increase in probability for syncing up, but not so weak that it has no effect on the probability of two coins syncing faces. For example, if the increase in probability for syncing up is 0.001%, the decay could be set to a value such as 0.0001% or 0.00005%.

Alternatively, you could consider using a non-linear function for the decay, such as an exponential decay, which would allow the probability of two coins syncing faces to decrease more rapidly as it gets further from 50%.

Ultimately, the best approach will depend on the specific goals of the model and the behavior that you are trying to achieve. It might be helpful to run simulations or do further analysis to explore different decay values and understand how they affect the behavior of the model.

Indeed. And there are even simpler starter sets or beginner modules than that one.

[Kane0]

A billion-dollar company like Hasbro isn't going to listen to the ideas of some random forum users, and neither is its Wizards of the Coast subsidiary. Not even if our ideas somehow gained wide internet traction. They're not even listening to the professionals they hired on purpose.

Furthermore, Wizasbro has broken and torn appart a trust built over decades, out of nothing but greed and condescension for the hobby's people.

Wizasbro was making money like no other RPG company ever made, with even its biggest competitor far behind. And they spat in our faces because to them it wasn't enough.

Maybe Wizasbro needs to adapt, but first it needs to stop the spitting and the spite and work to rebuild what they destroyed. The ball is in their court, but if that 30-pounds-a-month info is accurate, they still think they can use said ball to beat us up until we give them our lunch money.

This, but I will also attempt to be constructive.

- As I previously said, 'Steam for TTRPGs'. DMGuild is a start, incorporate and expand on it. Take a small cut from sales to generate that all-important exponential growth.
- Create/acquire and incorporate a tabletop tool. Even something as basic as Owlbear Rodeo with some elements syncing with the Beyond 'ecosystem', but include voice integration
- Create/acquire and incorporate writing tools. World Anvil is often touted but i dont have any personal experience with it, but something to help a DM in the creation process (and plug it into point 1 so it can be shared and monetised)
- Flesh out the encounter builder and have it be able to run mock battles for playtesting and simulation purposes, or let a bored user fight against AI.
- Throw in a mobile app for good measure, to access as many of these things as possible from phones and tablets so the tools are still useful for in-person games (and for people that do their best work while commuting on public transport)

Seeing a trend? All these things that mostly already exist. Bring it all in to make a single, cohesive, convenient package (preferably by not destroying the competition but lets not kid ourselves). Everything the community needs to make, share and play games in one place at a reasonable price. These are ways you can add value to the service instead of creating the illusion of it by undoing arbitrary limitations like number of characters you can save or drip-feeding unfinished or artificially delayed content that's already in the pipeline.
I personally would still hate an online service vs a product I can buy and own, but that ship has sailed so at least make it worthwhile.

[Atranen]

Spoiler: I have trouble believing something that can follow this conversation can't handle Lost Mine of Phandelver.

Show

Me
Imagine a fictional interpretation of fundamental reality that can be described with the metaphor of a field of coins. Each coin flips and takes a random, finite, non-instant amount of time to settle to either heads or tails, at which point the coin immediately flips again. If two adjacent coins happen to settle at the same instant with the same side facing up, there is a 0.001% chance those two coins will sync faces again on their next flip. This modification to probability is cumulative -- if they sync again, it compounds. If they don't sync, probability decays back toward 50% at the same rate. Do you understand this model at this point?

ChatGPT
I understand the model that you have described. It seems to be a way of representing the concept of probability, with coins flipping and settling in a certain position, and the probability of two coins settling in the same position at the same time affecting the probability of their future flips. Is there anything else that you would like me to clarify about this model?

Me
Okay, assuming an arbitrarily large field of coins, would the patterns of synchronized flipping eventually result in stable patterns of heads or tails?

ChatGPT
It is not clear to me whether the patterns of synchronized flipping in the model you have described would eventually result in stable patterns of heads or tails. In order for there to be stable patterns of heads or tails, the probability of two adjacent coins syncing faces on their next flip would need to be higher than 50%. However, in the model you have described, the probability of two adjacent coins syncing faces on their next flip only increases by 0.001% if they happen to sync on their current flip, which seems like it would not be enough to create stable patterns of heads or tails.

Additionally, the model you have described includes the idea that the probability of two coins syncing faces decays back toward 50% if they do not sync on their next flip, which could also work against the formation of stable patterns.

Without further information about the specifics of the model, it is difficult to say for certain whether stable patterns of heads or tails would emerge.

Me
Okay, let's change it so that the decay is only half as strong as the sync. So decay is 0.0005%. And don't forget that the increase in probability for syncing up is cumulative. Does that change your answer?

ChatGPT
With these changes to the model, it is still not clear to me whether stable patterns of heads or tails would emerge.

The decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up could potentially allow for the formation of stable patterns of heads or tails, depending on the initial probability of two coins syncing faces and the rate at which they continue to sync.

However, without further information about the specifics of the model, it is difficult to say for certain whether stable patterns would emerge. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Are there specific questions I could answer that could help you?

ChatGPT
There are a few things that could help me better understand the model you have described:
1: What is the initial probability of two adjacent coins syncing faces on their next flip? Is this probability the same for all pairs of coins, or does it vary in some way?
2: Is there a limit to the number of times that the probability of two coins syncing faces can increase?
3: Are there any other factors that could affect the probability of two coins syncing faces, such as the order in which the coins are flipping or the number of coins that are flipping at the same time?
Answering these questions might help me better understand the behavior of the model and how it might evolve over time.

Me
I'll answer by number. 1: Each coin begins with a 50% chance of heads or tails. The chance of any two coins syncing up is the same as it would be for any two real-world coins landing on the same side. It is the same for all pairs of coins, but we're only concerned with adjacent coins. 2: No limit to rate of increase. 3: The number of coins isn't directly affecting their flip rate.

ChatGPT
Understood. Given this information, it seems that the probability of two adjacent coins syncing faces on their next flip will remain relatively low, since it is based on the probability of two real-world coins landing on the same side, which is approximately 50%.

Additionally, the decay of the probability of two coins syncing faces back toward 50% being half as strong as the increase in probability for syncing up may not be sufficient to allow for the formation of stable patterns of heads or tails, since the initial probability of two coins syncing faces is relatively low.

It is possible that small, temporary patterns of heads or tails could form as the probability of two coins syncing faces increases, but it is unlikely that these patterns would be stable over time. The model you have described seems to involve a lot of randomness, so it is difficult to predict exactly how it will evolve.

Me
It seems like if we removed the decay entirely, that would significantly increase the likelihood of stable patterns over a large period of time.

ChatGPT
Yes, removing the decay of the probability of two coins syncing faces back toward 50% entirely would significantly increase the likelihood of stable patterns forming over a large period of time, compared to the version of the model you described where the decay is half as strong as the increase in probability for syncing up.

Without any decay, the probability of two coins syncing faces on their next flip would continue to increase cumulatively each time they sync, potentially leading to the formation of stable patterns of heads or tails.

However, it is still difficult to predict exactly how the model would evolve, since it depends on the initial probability of two coins syncing faces and the rate at which they continue to sync. It might be helpful to run simulations or do further analysis to better understand the behavior of the model under these conditions.

Me
Can you run a simulation like this?

ChatGPT
I'm sorry, but I am not able to run simulations. However, it is possible for you to write a program or script to simulate the model you have described and explore how it evolves over time. There are many programming languages that have libraries or built-in functions for generating random numbers or simulating probability, which you could use to implement the model.

Alternatively, you could try drawing out the model on paper and manually simulating it over a few iterations to get a sense of how it might behave. This could help you better understand the model and identify any patterns that emerge.

Me
Going back to the decay, could you suggest a relative strength for it that would reduce the inevitability of patterns forming but still be weak enough to encourage them?

ChatGPT
It is difficult to say exactly what strength the decay of the probability of two coins syncing faces should have in order to reduce the inevitability of patterns forming, but still be weak enough to encourage them, as this will depend on the specifics of the model and the desired behavior.
One possibility could be to set the decay to a value that is significantly weaker than the increase in probability for syncing up, but not so weak that it has no effect on the probability of two coins syncing faces. For example, if the increase in probability for syncing up is 0.001%, the decay could be set to a value such as 0.0001% or 0.00005%.

Alternatively, you could consider using a non-linear function for the decay, such as an exponential decay, which would allow the probability of two coins syncing faces to decrease more rapidly as it gets further from 50%.

Ultimately, the best approach will depend on the specific goals of the model and the behavior that you are trying to achieve. It might be helpful to run simulations or do further analysis to explore different decay values and understand how they affect the behavior of the model.

Spoiler

Show

This is a nice example, but I think we'd be better served by seeing an AI actually try to run Lost Mines. The model does a good job of acting like it knows what's going on, but the 'goldfish for memory' issue is still there. For example, even after you clarify (quite explicitly) "1: Each coin begins with a 50% chance of heads or tails. The chance of any two coins syncing up is the same as it would be for any two real-world coins landing on the same side. It is the same for all pairs of coins, but we're only concerned with adjacent coins." it continues to ask about initial conditions, e.g. "since it depends on the initial probability of two coins syncing faces and the rate at which they continue to sync."

As far as actually answering anything...everything it says is pretty vague.

I'm not rejecting the idea that an AI could run a d&d game successfully. But I haven't seen evidence that makes me believe it yet.

[stoutstien]

I just image an AI DM murdering most parties with LMoP as written without serious parameter prompts

[Kane0]

I can see them giving early access to the Core Rules digitally first (with the ability to purchase physical bundles cheaper with a subscription) and see how it goes for 6 months before sending the physical books to the printer in time for a holiday bundle. I also believe that some content such as shorter adventurers and whatnot will be exclusively digital. I can even see them doing something like a yearly bundle for the physical books with books released digitally throughout the year. In does make a lot of sense in regards to errata, typos, and avoiding Hadozee situations.

The plausibility of that horrifies me.