Jump to content

Recommended Posts

Posted

I am in the middle of Wind and Truth.

It has been years or reading these.  Thousands and thousands of pages.  I must admit a get lost or don't remember everything.  I have just wanted to ask questions like "Who is this herald again?" or "What was that character's back story again?" or "Is this the same unmade as we were talking about before?"

ChatGPT has been great.  I put it on speaker, tell it where I am at and that I don't want spoilers and have pretty lengthy conversations down all sorts of rabbit trails.

In Wind and Truth...

Spoiler

there is a scene with many of the heralds talking with each other.  I asked ChatGPT for a list of the heralds, orders, personalities, appearance, and final fate as we know it today.

It helped me understand the whole chapter in much more detail.

Anyone else using AI to sort out the deep lore?

Posted (edited)

Welcome to the Shard. Please consider an Intro Post to let us know what you have or have-not read (whichever list is shorter). Also, please consider checking out the Sharder FAQ for some useful forum info and tips. 

1 hour ago, owensbt said:

I am in the middle of Wind and Truth.

It has been years or reading these.  Thousands and thousands of pages.  I must admit a get lost or don't remember everything.

We also have the Coppermind to help explain and recall the details, if you need it. Please remember that Generative AI is stealing copyright material - and that is not a good thing. Brandon has a great article about AI and Art (written or otherwise).

Hope that helps

Edited by Treamayne
SPAG
Posted

I read that article.  Thanks for posting it.  I agree with it as well.  (My wife is an author) I want the humanity in my stories and am not looking for AI to replace that.

I do like using AI as a custom search engine to parse vast amounts of info and get me what I want. It is good at that. But more importantly, I value real people and real questions (and that is why I am here), even if that is slower and messier at times.  Real connection with real people are absolutely needed for art and in community.

So I think there is a time for both.

I think this is what my post was reaching for... The underlying motive was looking for connection with others on a new level of appreciation I have for the books when I can get clarification on things and keep reading.

Posted

I can sympathize with the idea but don't recommend it. I have tried using LLMs for some basic Cosmere research: I was trying to find a specific citation to post here, and knew it existed but didn't recall exactly where. I thought it might be faster to use the LLM to find it than to scan through the books manually. The LLM output was well organized, concise, specific, and 100% wrong. I followed up asking it to cite the sources for its wrong information, and it came up empty, confessing that the information provided did not match my question and the constraints I placed on it.

I followed up once again, asking for any reference to what I was looking for along with the relevant passage (if possible) and a book and chapter citation so that I could look it up myself. The results were, again, well organized, concise, specific, and 100% wrong. It cited the wrong chapter name (a chapter which doesn't exist in any SA or Cosmere book), the wrong chapter number for the information I wanted, an incorrect description of the information I wanted, and a wrong conclusion about what I was looking for. For reference, I was looking for something objectively answerable and specifically mentioned: the number of casualties during Desolations, and I was trying to find the vision in which Dalinar spoke with Nohadon after winning one because I recalled Nohadon giving a number (he did, plainly stated).

It will work, sometimes, but if the issue is that you're not sure you remember all of the relevant details then you won't be able to catch the errors and will get weird, garbled information which is likely to obscure more than it illuminates.

Posted
1 hour ago, Returned said:

I can sympathize with the idea but don't recommend it. I have tried using LLMs for some basic Cosmere research: I was trying to find a specific citation to post here, and knew it existed but didn't recall exactly where. I thought it might be faster to use the LLM to find it than to scan through the books manually. The LLM output was well organized, concise, specific, and 100% wrong. I followed up asking it to cite the sources for its wrong information, and it came up empty, confessing that the information provided did not match my question and the constraints I placed on it.

I followed up once again, asking for any reference to what I was looking for along with the relevant passage (if possible) and a book and chapter citation so that I could look it up myself. The results were, again, well organized, concise, specific, and 100% wrong. It cited the wrong chapter name (a chapter which doesn't exist in any SA or Cosmere book), the wrong chapter number for the information I wanted, an incorrect description of the information I wanted, and a wrong conclusion about what I was looking for. For reference, I was looking for something objectively answerable and specifically mentioned: the number of casualties during Desolations, and I was trying to find the vision in which Dalinar spoke with Nohadon after winning one because I recalled Nohadon giving a number (he did, plainly stated).

It will work, sometimes, but if the issue is that you're not sure you remember all of the relevant details then you won't be able to catch the errors and will get weird, garbled information which is likely to obscure more than it illuminates.

Same here. 

I've tried Googling specific topics to try and shortcut my way to reliable answers at times instead of searching the Arcanum directly, but Google AI also tries to answer the question itself. 

It draws on the many random ideas on the 17th Shard, Coppermind, and sometimes completely unrelated book series to confidently stitch together a totally incorrect Frankenstein of an answer.

 

AI can be helpful in finding links to more reliable sources (though many go to incorrect places), but in my experience it's really, really bad at coming to an accurate conclusion, especially on vague sci-fi/fantasy stuff.

Posted

Huh that's interesting, I found ChatGPT to generally give pretty accurate answers. Now I had to correct it a few times because it got Atium Feuchemy laughably wrong, but other than that it was pretty on point

Posted

An example from yesterday.  I am halfway through Wind and Truth and Thude was brought up.  I needed a memory jog for where I last saw him and a bit about his story.  ChatGPT helped.

I also was listening to a chapter referencing all the heralds by name and I was trying hard to remember who was with what order and which we had met already in the "present".  A quick search and I had a reference chart that helped me enjoy the chapter instead of just being confused.

As @Treamayne pointed out, LLMs should not be able to just read the book.  I have found that if I ask what chapter something happened in, that it does a pretty poor job.  I am okay with that.  Perhaps that is because they appropirately don't have direct access so they can't answer super specific questions on where something was referenced.  They should have to pay for the book if they are going to have direct access to that.

Posted
15 minutes ago, owensbt said:

An example from yesterday.  I am halfway through Wind and Truth and Thude was brought up.  I needed a memory jog for where I last saw him and a bit about his story.  ChatGPT helped.

I'm not certain how long that kind of query takes, but one word in Coppermind finds Thude - with all relevant citations. Hard to think it is easier than that. Actually, with Coppermind autocomplete, you sometimes don't even need to type a full word to find the info you want. . . and since you have access to the Time machine, you can also avoid spoilers for books after the one you are reading. 

Hope that helps

 

Posted

Anything that's got an entry in Coppermind or the Arcanum will probably be pretty reliable (up to the most recent training date), though not 100%. That kind of information is unlikely to cause confusion, though I agree with @Treamayne that searching those sources yourself is not a heavy lift (and it will be more reliable). There isn't much question that the LLMs have scraped Coppermind, along with its excellent linking and citations, so that's not surprising.

Going beyond something that is plainly written in one of those spots can quickly become problematic, and while that's easy to fall into the types of things @owensbt is describing probably won't dip into that territory. I'm not sure I'd trust it to shield me from spoilers but that's a risk you run no matter what you do.

As a side note, many LLMs almost certainly have consumed all of these books (it's been acknowledged by many of the companies that run them that they trained on a huge amount of stolen copyrighted material, including published books). I doubt they'll return much about those books' content, for now, but if they did it wouldn't be that different from dedicated fans collating information on a site like Coppermind.

Posted
9 minutes ago, Returned said:

I doubt they'll return much about those books' content, for now, but if they did it wouldn't be that different from dedicated fans collating information on a site like Coppermind.

Except for the whole "We received permission and follow the Guidelines set forth by Team Dragonsteel regarding content and artwork"
 - Also why we do not reproduce whole sections of book - and summaries are made by the editors (e. g. I wrote most of the White Sand Summary)

 

I am not saying others are wrong for using a tool (each person has the right to decide for themselves) - I am only saying that I cannot, personally, morally justify using a tool based on stolen Intellectual Property. 

Posted (edited)
14 minutes ago, Treamayne said:

Except for the whole "We received permission and follow the Guidelines set forth by Team Dragonsteel regarding content and artwork"
 - Also why we do not reproduce whole sections of book - and summaries are made by the editors (e. g. I wrote most of the White Sand Summary)

Sorry, I wasn't clear-- I didn't mean it's ethically the same thing, more that it's mechanically very similar.

"I didn't understand [X] in this chapter of this book, can someone clear it up for me?" is a question someone might ask here, and get a summarized answer from readers who participate on the forum (including making mistakes!), along with cited excerpts. It doesn't come up (because no one asks, because it doesn't really matter) if those readers borrowed that book from the library or a friend or bought it from a used bookstore or garage sale, or even stole it. LLMs trawling through Coppermind, the Arcanum, the 17th Shard, subreddits, and similar to aggregate all of that second-degree information is an interesting situation, even if the LLM has never consumed any of the books.

Edited by Returned
Posted (edited)

I like using the coppermind (as was recommended by @Treamayne) to give me chapter summaries. It can be especially helpful if a day or so goes by, and I just want a concise refresher for the past couple chapters, such as I've fallen asleep reading, and the next day I had slightly forgotten what I'd read.

The AI route with ChatGPT, Gemini, and Brave's AI seem to be work, but not all the time and not to any great effiency. Sometimes just mentioning that you want only a review of a certain chapter and not to spoil with chapters after it will result in a complete spoiler-free review of that chapter with a summary such as "Shallen is facing an issue and needs to do it her way or follow another's advice... She realizes that she is growing, but she wonders if the cost is worth it... The time has come and she knows she can't keep on depending on her friends to bail her out..." Essentially, just a spoiler-free, vague, outline, and not at all what I asked for.

That said, sometimes it will do what you want, somewhat.

Currently, I'm reading Oathbringer, about 50 chapters in. I decided to ask ChatGPT to do a summary of chapter 49 with "give me a chapter summary of chapter 49 of oathbringer by brandon sanderson"

It gave a decent result this time:

Spoiler

In Chapter 49, the point of view switches to Dalinar Kholin in a flashback that takes place twenty-three years before the main events of Oathbringer. Dalinar, restless and craving the chaos of battle, finds himself in a tavern with his men. He’s surprised when they all refuse to fight him — not out of cowardice, but because Dalinar, while intoxicated in a previous brawl, beat three men so severely that they were permanently crippled. News soon arrives that Dalinar’s wife, Evi, has given birth. After meeting his newborn son, Dalinar talks with his brother-in-law and then with Gavilar about the ongoing rebellion at the Rift. Gavilar’s diplomatic efforts have failed, and he sends Dalinar off to command border skirmishes, hoping Dalinar’s martial prowess will both serve Alethkar and demonstrate its strength to its enemies.

I found the coppermind summary had given much more information which really helped to remind me what happened:

Spoiler

Dalinar uses some firemoss to escape his quiet life with no battles or fighting. Havar and Bashin are with him and place bets on the wrestling match. Dalinar watches the match and begins to feel excited at the conflict, then he remembers that that almost led him to kill Gavilar. Havar asks when they are going to the Rift to fight the rebellious Tanalan, but Gavilar is trying diplomatic means. The feature match has been canceled due to one participant being sick, so Dalinar takes off his shirt and enters the ring. His opponent is the undefeated Makh, but he backs out for fear of being hurt. Havar explains that Dalinar maimed some men in a tavern fight a couple of weeks back, so no one wants to face him. The den empties and Toh arrives, looking for Dalinar. He reports that Evi is giving birth.

Evi tells Dalinar that he has a son, and the midwives bring him in. He takes him in his arms and laughs in joy. He gives him a silent blessing: that he may inherit his strength and some of Evi's compassion. Looking at his son, he better understands Gavilar's thought for a stable future. He names him Adolin and is surprised to see Gavilar there. He congratulates Dalinar, but there is clearly something wrong, so they go apart to talk. Dalinar asks after Jasnah's "lunacy," but he says she's recovering. Instead, he's worried about the Rift, which is in true rebellion. Dalinar worries about going back to battle because he's an animal when he fights. Gavilar instead sends him to fight some other border disputes to show that they are determined to keep a united Alethkar.

The mention of Dalinar using a particular substance was what really triggered me to remembering what happened, due to the addictions which another character had in a previous chapter. ChatGPT didn't think to mention it.

Gemini, for comparison, printed out some sort of study note bullet points and other info that sounded more like a paper a student would hand into a teacher. It wasn't at all preferable to read as a summary as it wasn't nearly as efficient:

Spoiler

In Chapter 49 of Oathbringer, titled "Total Commitment," the focus shifts to the intense political and personal maneuvering in Urithiru as Dalinar Kholin attempts to forge a coalition against the Voidbringers.

Key Plot Points
The Vision Strategy: Dalinar decides to bring world leaders into his visions to prove the True Desolation is happening. In this chapter, he brings Fen Rysn (the Thaylen Queen) and her advisors into a vision of Aharietiam—the aftermath of a Desolation.

The Harsh Reality: The vision is brutal. It shows the sheer scale of destruction and the fact that the Heralds abandoned their posts. Dalinar’s goal isn't just to scare them, but to show that humanity can survive if they unite.

Fen’s Skepticism: Queen Fen is a pragmatist. She is deeply suspicious of Dalinar, fearing that his "unification" is just a cover for Alethi conquest. She challenges him, noting that the Alethi have a history of being "the girl who cried Voidbringer" to justify war.

The Navani Factor: Navani plays a crucial role as the diplomatic anchor. While Dalinar provides the raw, visionary power, Navani manages the optics and the technical details of the coalition, ensuring the other monarchs feel like partners rather than subjects.

The "New" Dalinar: We see Dalinar struggling with his past as the Blackthorn. He wants to be a man of peace and philosophy (following The Way of Kings), but he realizes that to lead, he must still possess the iron will of a warlord.

The Emotional Core
The chapter highlights Dalinar’s desperation. He is fighting a war on two fronts: one against the Fused, and one against the reputations of himself and his nation. He realizes that trust is a harder currency to earn than spheres.

Significant Character Moments
Dalinar Kholin: Shows growth by refusing to take offense at Fen's insults, choosing instead to be vulnerable about his past mistakes.

Queen Fen: Establishes herself as a formidable political opponent who won't be easily swayed by magic tricks or divine claims.

"I’m not looking for a conquest, Your Majesty. I’m looking for a way to keep us all from becoming a memory." — Dalinar's sentiment during the meeting.

As you can see, most of that is pure garbage... Not to mention that it's not even describing chapter 49. It's another chapter entirely.

We know that coppermind summaries will give a concise recap, true information, and no spoiling, so why not just use it?

Edited by bedtime
Posted
2 hours ago, Treamayne said:

since you have access to the Time machine

I didn't know about that.  I will look at it.  

Posted

I am hearing the support for Coppermind and appreciate it and see its value.  Both as an archive and for active conversation.

What is still like about chatgpt is the casual ease.  I hit one audio button on my phone and just talk.... Even dialogue.  It is more like this:

-----

Wait... I just read this ... Is this the same person that did x back in another book.

"No... But I see why you are confused.  That person was y and the name sounds similar.  This is x from this context"

Oh...I think I see something... Are they connected with this larger theme of this over in mistborn

"Well.. that hasn't been spelled out, but on fan sites many have speculated about that"

Ah .. so I not missing something I am supposed to know for now.

"No... You have it. This is new and you know what you are supposed to know at this point."

Cool.  Back to reading.

---

So different tools for different uses.

 

Posted (edited)

Personal opinion

Spoiler

That sounds messed-up

(Having a conversation with an AI like, over the phone)

Edited by Usseewa
Posted
49 minutes ago, owensbt said:

I am hearing the support for Coppermind and appreciate it and see its value.  Both as an archive and for active conversation.

What is still like about chatgpt is the casual ease.  I hit one audio button on my phone and just talk.... Even dialogue.  It is more like this:

-----

Wait... I just read this ... Is this the same person that did x back in another book.

"No... But I see why you are confused.  That person was y and the name sounds similar.  This is x from this context"

Oh...I think I see something... Are they connected with this larger theme of this over in mistborn

"Well.. that hasn't been spelled out, but on fan sites many have speculated about that"

Ah .. so I not missing something I am supposed to know for now.

"No... You have it. This is new and you know what you are supposed to know at this point."

Cool.  Back to reading.

---

So different tools for different uses.

 

I am fine with you having your own opinions, but I hope you understand the environmental and ethical impacts of AI. AI companies destroy the environment constantly to get out incorrect stolen information. There is a great video on AI right here, that I think explains the ethical and environmental impacts of AI well.

After that, if you still decide to use AI, I understand that to be your choice as an informed human.

Posted
3 hours ago, Usseewa said:

Personal opinion

  Hide contents

That sounds messed-up

(Having a conversation with an AI like, over the phone)

realistically, I don't find it any different from typing to an AI, which, if one is treating the AI like a friend rather than a tool, I find weird, but otherwise I think it's fine. It's just like voice searching Google, except you can ask follow-ups easier

Posted

I was making a Veden character for a Stormlight RPG game I'm playing in. I found the list of Veden names in the World Guide to be extremely lackluster, and Category:Veden on the Coppermind was also horrifically small, with just around forty characters. It wasn't really enough for me to fully understand the conventions and create two Veden noble house names.

So, I thought hey, maybe this is a thing ChatGPT would be good at, synthesizing a bunch of stuff to give me some points to jump off from. I asked it for more Veden names... 

"Here’s a list of original Veden-style surnames I’ve crafted for your RPG character, inspired by canonical names like “Sadeas,” “Torol,” and “Meridas”"

Yeah... no. Those aren't Veden, ChatGPT! 

Any time I try AI--which is rare--I feel like it immediately face plants. (Admittedly, it's not like there's a lot of Veden source material for it to draw upon.) 

Posted
1 hour ago, Chaos said:

I was making a Veden character for a Stormlight RPG game I'm playing in. I found the list of Veden names in the World Guide to be extremely lackluster, and Category:Veden on the Coppermind was also horrifically small, with just around forty characters. It wasn't really enough for me to fully understand the conventions and create two Veden noble house names.

So, I thought hey, maybe this is a thing ChatGPT would be good at, synthesizing a bunch of stuff to give me some points to jump off from. I asked it for more Veden names... 

"Here’s a list of original Veden-style surnames I’ve crafted for your RPG character, inspired by canonical names like “Sadeas,” “Torol,” and “Meridas”"

Yeah... no. Those aren't Veden, ChatGPT! 

Any time I try AI--which is rare--I feel like it immediately face plants. (Admittedly, it's not like there's a lot of Veden source material for it to draw upon.) 

Take any word in Hebrew and just tweak it and then u get a nice Veden name...

Gerev (sock) -> Gerev, Gereven, Gerevan, Gereval
Bet sefer (school) -> Batsef
Shoresh (root) -> Ioresh, Ioreshar, Moresh, Toresh... add -ar or -an for good measure

Anan (cloud) -> Ananat, Ashnan

sometimes words on their own work too: Anan (cloud), Beten (tummy), Delet (door), Devash (honey), Hodesh (month)

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

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