Quote from: HappyDaze on March 15, 2025, 03:04:10 PM

Quote from: RNGm on March 15, 2025, 02:36:47 PM

Quote from: HappyDaze on March 15, 2025, 09:11:44 AM

Quote from: Fheredin on March 15, 2025, 08:17:23 AMChances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

Do you have any evidence to support this claim?

Anyone can generate the evidence in no time flat.       AI would have to be purposefully and consistently mistrained like the diverse and female Catholic pope and founding father generating google AI gemini in order to have to come up with the 2024 art at least.

So that's a "no" then?

It was a joke.  I made no claim... wrong person.

Quote from: RPGPundit on March 15, 2025, 03:57:27 AM

Quote from: Man at Arms on March 13, 2025, 05:22:57 AMWOTC will never beat 2014 5E D&D.  That shadow will forever loom large, over them.

Well, they certainly won't beat it if they don't hire actual game designers. But I could see someone someday beating it, whether that's WotC or some company they license D&D to.

Pundit, I was really saying that WOTC specifically, can't and won't beat it.  The don't have the personnel to do it.  Their pride won't allow them to admit their mistakes, and hire real game designers.

A good design team and good market timing, might beat 2014 5E; but it's not a given.  What a money maker, it was.

Quote from: Omega on March 15, 2025, 05:49:01 AM

Quote from: yosemitemike on March 03, 2025, 12:05:14 AMI think part of it is timing.  WotC is trying to move D&D to a live service subscription model with microtransactions at the same time that many consumers are soundly rejecting that model.  A lot of people are just fed up with this sort of thing. 

Alot of venues are sometimes as much as a decade out of touch with their customer base.

Others were never in touch with their customer base.

This is why I don't travel 1 hour to my nearest game store anymore.  They don't want my business.  I used to buy dice just to patronize them, when they didn't have a single thing I wanted.  But they made it apparent that they don't want business from people who think differently than they do.

Quote from: RPGPundit on March 15, 2025, 03:57:27 AM...
Well, they certainly won't beat it if they don't hire actual game designers. But I could see someone someday beating it, whether that's WotC or some company they license D&D to.

Not only will somebody someday beat it, but the bar is not as high as people think.

Quote from: HappyDaze on March 15, 2025, 09:11:44 AM

Quote from: Fheredin on March 15, 2025, 08:17:23 AMChances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

Do you have any evidence to support this claim?

The Golaith's Powerful Build feature does not work properly RAW. It gives you advantage on saving throws to end being Grappled, but you don't end Grappled with a saving throw. You end it with an Athletics check.

This is an example of a mistake which a human could have made, but the idea that a human WotC employee actually tasked with writing rules would have not cross-checked how Grappling works is a bit ridiculous.

However, an LLM is a probability sieve where it interpolates the next most probable word. Being grappled is a condition, and the most likely next word to be associated with a condition is a saving throw. This is a mistake a human could make, but probably wouldn't...but it is also a mistake an LLM would make most of the time. This is an example where there is a significant difference between a human comprehending the words and the LLM computing the next token with a probability calculation.

It's my educated guess that some sections of the 2024 PHB were written by LLMs and then proofread by humans, and because proofreading doesn't have great odds to catch flaws like this, some of the mistakes the LLM made escaped all the way to production.

Quote from: Fheredin on March 16, 2025, 09:09:34 AMThe Golaith's Powerful Build feature does not work properly RAW. It gives you advantage on saving throws to end being Grappled, but you don't end Grappled with a saving throw. You end it with an Athletics check.

This is an example of a mistake which a human could have made, but the idea that a human WotC employee actually tasked with writing rules would have not cross-checked how Grappling works is a bit ridiculous.

FWIW, that type of mistake is pretty common in games.  It could be a simple human mistake as you mentioned or, even equally likely, could be the result of a change that wasn't fully carried over to all references to the test during development.   At one time, they may have been testing escaping a grapple with a strength save and decided not to ultimately but forgot to change it in the goliath rules where it was mentioned.

Quote from: Fheredin on March 16, 2025, 09:09:34 AM

Quote from: HappyDaze on March 15, 2025, 09:11:44 AM

Quote from: Fheredin on March 15, 2025, 08:17:23 AMChances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

Do you have any evidence to support this claim?

The Golaith's Powerful Build feature does not work properly RAW. It gives you advantage on saving throws to end being Grappled, but you don't end Grappled with a saving throw. You end it with an Athletics check.

This is an example of a mistake which a human could have made, but the idea that a human WotC employee actually tasked with writing rules would have not cross-checked how Grappling works is a bit ridiculous.

That's where you're absolutely wrong. The human beings currently employed by WotC are MONUMENTALLY incompetent.

How bad? We know this from the Book of Many Things: Mackenzie de Armas has no idea how to write RPG rules, and lacks the mental capacity to ever learn, so she took full credit for the product but it was actually almost entirely written (except for Armas' Lesbian Fanfiction) by Jason Tondro.
And TONDRO SUCKED TOO.

He's one of the last people left at WotC who has any knowledge of game design, and under the circumstances of a more normal hobby he should probably be completely incapable of finding any job above that of mailroom boy.

The biggest proof that this edition WASN'T written by AI is that if it was true the game would probably be a lot better.

Quote from: Fheredin on March 16, 2025, 09:09:34 AM

Quote from: HappyDaze on March 15, 2025, 09:11:44 AM

Quote from: Fheredin on March 15, 2025, 08:17:23 AMChances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

Do you have any evidence to support this claim?

The Golaith's Powerful Build feature does not work properly RAW. It gives you advantage on saving throws to end being Grappled, but you don't end Grappled with a saving throw. You end it with an Athletics check.

This is an example of a mistake which a human could have made, but the idea that a human WotC employee actually tasked with writing rules would have not cross-checked how Grappling works is a bit ridiculous.

However, an LLM is a probability sieve where it interpolates the next most probable word. Being grappled is a condition, and the most likely next word to be associated with a condition is a saving throw. This is a mistake a human could make, but probably wouldn't...but it is also a mistake an LLM would make most of the time. This is an example where there is a significant difference between a human comprehending the words and the LLM computing the next token with a probability calculation.

It's my educated guess that some sections of the 2024 PHB were written by LLMs and then proofread by humans, and because proofreading doesn't have great odds to catch flaws like this, some of the mistakes the LLM made escaped all the way to production.

IOW, you're pulling a guess out of your ass.

It's also hilarious to me that you say a human writer would be unlikely to make this kind of mistake and also unlikely that a human proofreader would be unlikely to catch it. Why would you think this? As someone that does proofreading, I can tell you I see writers make mistakes like this all the time, and I catch them all the time.

Quote from: HappyDaze on March 16, 2025, 01:00:48 PM

Quote from: Fheredin on March 16, 2025, 09:09:34 AM

Quote from: HappyDaze on March 15, 2025, 09:11:44 AM

Quote from: Fheredin on March 15, 2025, 08:17:23 AMChances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

Do you have any evidence to support this claim?

The Golaith's Powerful Build feature does not work properly RAW. It gives you advantage on saving throws to end being Grappled, but you don't end Grappled with a saving throw. You end it with an Athletics check.

This is an example of a mistake which a human could have made, but the idea that a human WotC employee actually tasked with writing rules would have not cross-checked how Grappling works is a bit ridiculous.

However, an LLM is a probability sieve where it interpolates the next most probable word. Being grappled is a condition, and the most likely next word to be associated with a condition is a saving throw. This is a mistake a human could make, but probably wouldn't...but it is also a mistake an LLM would make most of the time. This is an example where there is a significant difference between a human comprehending the words and the LLM computing the next token with a probability calculation.

It's my educated guess that some sections of the 2024 PHB were written by LLMs and then proofread by humans, and because proofreading doesn't have great odds to catch flaws like this, some of the mistakes the LLM made escaped all the way to production.

IOW, you're pulling a guess out of your ass.

It's also hilarious to me that you say a human writer would be unlikely to make this kind of mistake and also unlikely that a human proofreader would be unlikely to catch it. Why would you think this? As someone that does proofreading, I can tell you I see writers make mistakes like this all the time, and I catch them all the time.

So, basically, his guess comes from the same place that every comment you make on this website comes from.  Not exactly the own you think it is...

Quote from: RPGPundit on March 16, 2025, 12:07:39 PMBook of Many Things

Well, I'm glad to say that I'm now so oblivious to what WotC is doing that I had no idea what this was. Apparently they've released a number of books since I stopped paying attention (and I used to buy every new release in the early days of 5E!). Now I wouldn't touch a WotC book with a ten-foot pole.

QuoteThe biggest proof that this edition WASN'T written by AI is that if it was true the game would probably be a lot better.

This was my thought as well. AI can do incredible things. WotC, not so much.

You know, it is possible for a gross incompetent using poor AI to rapidly produce something worse than either could working alone.  Just saying.

Quote from: Steven Mitchell on March 16, 2025, 02:33:18 PMYou know, it is possible for a gross incompetent using poor AI to rapidly produce something worse than either could working alone.  Just saying.

True. We're in the "dawizard" phase of AI: an amazing cost-saving tool for those who know how to use it properly, a tool of mass destruction and embarrassment for those who don't.

As we all know what you're responding to, I'll rearrange things thusly. I do have experience on this forum dogpiling thing.

Quote from: RNGm on March 16, 2025, 10:36:46 AMFWIW, that type of mistake is pretty common in games.  It could be a simple human mistake as you mentioned or, even equally likely, could be the result of a change that wasn't fully carried over to all references to the test during development.  At one time, they may have been testing escaping a grapple with a strength save and decided not to ultimately but forgot to change it in the goliath rules where it was mentioned.

Quote from: RPGPundit on March 16, 2025, 12:07:39 PMThat's where you're absolutely wrong. The human beings currently employed by WotC are MONUMENTALLY incompetent.

How bad? We know this from the Book of Many Things: Mackenzie de Armas has no idea how to write RPG rules, and lacks the mental capacity to ever learn, so she took full credit for the product but it was actually almost entirely written (except for Armas' Lesbian Fanfiction) by Jason Tondro.
And TONDRO SUCKED TOO.

He's one of the last people left at WotC who has any knowledge of game design, and under the circumstances of a more normal hobby he should probably be completely incapable of finding any job above that of mailroom boy.

The biggest proof that this edition WASN'T written by AI is that if it was true the game would probably be a lot better.

Quote from: HappyDaze on March 16, 2025, 01:00:48 PMIOW, you're pulling a guess out of your ass.

It's also hilarious to me that you say a human writer would be unlikely to make this kind of mistake and also unlikely that a human proofreader would be unlikely to catch it. Why would you think this? As someone that does proofreading, I can tell you I see writers make mistakes like this all the time, and I catch them all the time.

What all three of you miss is the anatomy of the workflow logic that led to the mistake. For a really complex exception-based rules system like D&D, a human writer will probably have a second screen with the developer equivalent of the SRD open, so if you write an ability, you will Ctrl + F to the rules in question. If you don't do this, you are literally going to be making mistakes all the bloody time, regardless of how familiar with the system you are. This means that the more odd the rules the SRD is going to use, the less likely a human author is going to make a mistake. It also means that paradoxically, the less familiar with the system the author is, the less likely they are to make a mistake because they will lack the confidence to just write the rules and will instead double-check before writing.

It's when you start dealing with more mundane rules or rules writers who are more confident with the system that you really need to watch for mistakes, because these are the situations people won't double-check a mistake in.

Now the interesting part is that most LLMs do not necessarily double-check their work the way a human would. It's also worth noting that the 2024 D&D PHB was probably written in 2023, which was well before self-prompting and thinking models like o1 Strawberry (which effectively released the same time the PHB did, although I think it's doubtful a thinking model would have avoided this mistake.) It's also worth noting that this was around the time of WotC's first layoff, but not the more recent ones. If you seriously think that AI would be better than human, I think that's because you're either poo pooing WotC in general or not accurately assessing what they actually had to work with when they were actually writing the book.

Proofreading (and playtesting, which I don't think anyone actually thinks WotC did at any real scale) is almost certainly going to favor putting a fine detailed eye over abilities which threaten game balance, which means that it will pay a lot more attention to casters than to melee. This is the kind of mistake which will escape general proofreading, and clearly missed internal playtesting, purely because of where these people will spend the majority of their time when reading the book.

The last thing to note is that, like I said in my first post on this, this is exactly the kind of mistake you would expect an LLM to make. It sees condition, it writes advantage on saving throw because these are all words it is expecting to string together.

Quote from: Fheredin on March 15, 2025, 08:17:23 AMIf that's the case for Magic--the goose which lays the golden eggs--then the 2024 PHB was likely a simple cash grab. Chances are large stretches of the book were AI generated and WotC may have limited intention to give it long term support.

I doubt fake 5e will last even 5 years before wotc comes out with BETTER! fake 5e 3.8! You'll create characters using real cookie cutters and instead of dice you will just play rick paper scissors because "modern" players cant understand complex things like 1+1. And all this is TOTALLY backwards compatible!

Quote from: Steven Mitchell on March 16, 2025, 02:33:18 PMYou know, it is possible for a gross incompetent using poor AI to rapidly produce something worse than either could working alone.  Just saying.

Seen this a LOT outside of the game-design space, in places where lives are on the line daily. And people with paper-credentials with very little experience using AI, don't *really* know their fundamentals in practice, spitting out dogshit that can actually cause things to fail and hurt people.

AI will replace those people, even in the game-space, as it raises the median bar. People that know how to interact with it, that have good fundamentals in their craft (whatever that is), will only get better with experience using AI. All bets are off if/when AGI drops.