Please read. Significant change on the site that will affect compatibility [ Dismiss ]
Home ยป Forum ยป Author Hangout

Forum: Author Hangout

ChatGPT as poor man's developmental editor

JoeBobMack ๐Ÿšซ

I'm interested in how AI might be used to help amateur authors improve (and, yes, I am one!). As likely everyone knows, it's super hard for beginners to analyze their own works, and even experienced authors can be too close to be objective. I know tools exist that can do some of this for a fee, such as Marlowe at authors.ai, but I haven't tried them. I was curious to see what the new version of ChatGPT with DataAnalysis could do, so I obtained a PDF of Princess of Mars by E.R. Burroughs, dropped it in, and asked for a 500 word summary, a character list, a description of the relationship of John Carter and Deja Thoris, and a psychological profile of John Carter. It's been a LONG time since I read the book, but everything I got back seemed pretty reasonable. I mentioned the Myers-Briggs Type Indicator in the last prompt, and that's what I got.

Questions:

What are others trying like this with the new ChatGPT? How close can ChatGPT, at $20/month, come to being, say, a poor man's developmental editor, or at least constructively critical reader?

What is possible in this area? OpenAI has announced custom "gpts." I'm wondering if other text analysis tools could be utilized by a "gpt" -- sort of like the way Zapier can link multiple apps -- to do even more?

What should authors be concerned about in using ChatGPT with their works-in-progress?

If anyone would like for me to take your story from here and run it through what I did with A Princess of Mars, let me know.

Quasirandom ๐Ÿšซ

@JoeBobMack

Developmental editor? Not very good. One way to think about it is that most generative AIs have a very short attention span, and can't remember details over long distances.

But as a copy editor/proofreader, the one who cleans up sentences and straightens out paragraphs, they can work pretty well. You have to train them or give them guidance on what your target style is, but once you do that, they can do a decent job.

Gauthier ๐Ÿšซ

@JoeBobMack

What should authors be concerned about in using ChatGPT with their works-in-progress?

Authors can't use chatGPT with the current term of use.

https://open.ai/terms

By using Open AI website and services, you hereby agree to forfeit all intellectual property rights claims worldwide, and forfeit any/all copyright claim(s), to the Content you provide or receive through your use of Open AI website and services, regardless of legal jurisdiction or intellectual property law applicable therein

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Gauthier

Gauthier, thanks for pointing that out. And, it raises a REALLY weird issue, because OpenAI is (a) inviting users to submit their individually-owned work for analysis by the Creative Writing Coach gpt that they've put out and (b) they have contradictory Terms of Service! See https://openai.com/policies/terms-of-use dated March 14, 2023 which states:

3. Content
(a) Your Content. You may provide input to the Services ("Input"), and receive output generated and returned by the Services based on the Input ("Output"). Input and Output are collectively "Content." As between the parties and to the extent permitted by applicable law, you own all Input.

That seems WAY more reasonable, not that "reasonable" and "Terms of Service" are necessarily compatible concepts!

I'm curious enough about this that I've put a post on OpenAI's subreddit and will update here if anything significant comes of it.

Replies:   JoeBobMack  JoeBobMack
JoeBobMack ๐Ÿšซ

@JoeBobMack

Actually, I think I've got it. Open.ai is a site that only accepts input in the form of prompts and outputs images. Openai.com is the company's site and it links to chat.openai.com which then links to the the ChatGPT tool. So, the Terms of Service for openai.com are applicable. Which only makes sense as the company is inviting users to upload data to use in private GPTs. So it only makes sense that OpenAI recognizes the ownership rights of its users. To do anything else would be like Microsoft claiming rights to anything produced with Word! (I'm not saying they wouldn't if they could, just that no one would ever use their products!)

JoeBobMack ๐Ÿšซ

@JoeBobMack

OpenAI has new Terms of Use that show up when you use ChatGPT but state they apply to other tools like DALL-E. These terms are effective December 14,2023, and they indicate that the previous Terms were the March, 2023 version mentioned earlier. So, I think that the fear that supplying content to OpenAI wil constitute putting that content into the public domain can be laid to rest. You own it after just as before. See the Terms of Uervice here.

Replies:   DBActive
DBActive ๐Ÿšซ

@JoeBobMack

Those are only EU terms. The US terms differ.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@DBActive

I don't think so, though it doesn't seem to matter. Here is the link to the EU Terms: https://openai.com/policies/eu-terms-of-use. The content ownership terms appear identical.

Replies:   DBActive
DBActive ๐Ÿšซ

@JoeBobMack

From your link:

If you reside in the European Economic Area, Switzerland, or the UK, your use of the Services is governed by these terms.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@DBActive

I know. But "these terms" is a hyperlink that takes you to a page that is clearly the EU Terms. And, as to user owned content uploaded to openai.com sites, the teens are the same: user retains all rights.

Gauthier ๐Ÿšซ

@JoeBobMack

Even with an AI service where the generated content copyright would be assigned to the user, using such output would currently void the copyright on that part in the US.

Note that this is an US restriction, publishing in another country like France, the copyright would be valid and the US would be forced to honour it.

See:
https://crsreports.congress.gov/product/pdf/LSB/LSB10922

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Gauthier

generated content copyright

Is that applicable here? The user is applying the service to content the user already created and this holds the copyright for. The user would Not be selling to generate a copyright for material generated by ChatGPT, correct?

Well, I guess if the author asked for wholesale changes the speed then verbatim, maybe, for those portions. Though how that would ever come up in a copyright infringement case, I can't imagine.

Replies:   Gauthier
Gauthier ๐Ÿšซ

@JoeBobMack

With the current terms, by submitting the text to ChatGPT, it is made public domain. The only recourse for the author would be asserting that the person doing the prompt stole the content without consent from any copyright holder.
In that case, openAI could require indemnification by the prompter.

Previous terms assigned the copyright to openAI then it was changed to the user, now it is to the public domain.

A side benefit for OpenAI is that when their server will be hacked and all Prompt/Answer leaked, you would have no recourse against OpenAI because the leak only contained content made public by the license terms.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Gauthier

Interesting. So you think the language from open.ai is controlling? I just don't know, but I can imagine the discovery that would go on should that ever be litigated. However, I have an extraordinarily difficult time imagining how that would come up. On the other hand, I'm shocked, shocked I tell you, that their legal team has let this situation of conflicting published terms of service ever arise! Bad lawyers, bad!

Switch Blayde ๐Ÿšซ

@JoeBobMack

I don't want artificial intelligence guiding my work.

I'm writing for people with real intelligence.

(said flippantly with a lot of sarcasm and a little truth)

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Switch Blayde

Yeah, the author has to supply the true intelligence. Other than that, it strikes me as not that much different than the tools in Word, ProWritingAid, Grammarly, etc. -- just a bit more sophisticated.

I don't think you would criticize an author who hired an editor and thereby learned of a flaw in their writing that they could improve. If an AI tool helps authors without those kind of resources but with the drive and will to improve to get better and produce more enjoyable stories, then to me we all win!

JoeBobMack ๐Ÿšซ

@JoeBobMack

Well, at the risk of having lost my copyright (I'm not that worried), I submitted a sizable (20,439 words) first chunk of the first book in a series I've been working on to the ChatGPT Creative Writing Coach ("Coach"). My understanding is that the current limits of Coach would allow a much larger chunk, maybe over 60,000 words, but I'm not positive about that.

I also apologize that I didn't make a copy of my prompt to the Coach before I deleted the conversation. However, I said something like that it had been asked as a developmental editor to review the material and provide an analysis, including a summary of approximately 500 words. Here's the summary:

**Summary:**
The text opens with a prologue that speaks to the magical community, acknowledging the need for a more comprehensive understanding of their world, which has evolved from a simple guide called "Magical Sex Slavery for Dummies" into a series of books starting with "Fate's Agent." It outlines the readers' requests for deeper insights into magic, relationships, and personal stories.

The novel then shifts to a character named Daniel, who, after an encounter with a goddess, is transported back in time to the town of Corden, Alabama, in 1973. His mission is to quickly settle in this era and take his first slaves to mitigate risk, a concept he struggles with. He meets Diana, a woman who disrupts Sheriff Jefferson's suspicious questioning of him. Diana, intrigued and disenchanted with her life, drives him away from the Sheriff, despite her disbelief in Daniel's story about being a student working on a project.

In a turn of events, Diana, who is grieving her past mistakes and her husband's death, feels an inexplicable connection with Daniel and agrees to take him to her home. Here, Daniel promises to share his story and give her choices that could potentially change her life. The narrative then reveals Daniel's backstory, including his visit to his wife's grave and his encounter with the goddess.

That's spot on. Ok, ok, I might quibble a bit with "inexplicable," but I admit things move fast after they meet, though for reasons I think make sense as they come out in the course of the story.

Anyway, defensive justification of my writing aside, it then had mostly complimentary things to say about the characters, plot, dialogue, and setting. Maybe it's just overly positive, but it's comments weren't all that dissimilar to feedback I've gotten from editors on this site. I didn't dig deeper into these, but maybe that will be a future effort. It would be interesting to have it analyze my main character and see how well it tracked the personality and role of that character in my mind.

Finally, I had asked about marketability, and it said:

**Marketability:**
The novel's unique blend of fantasy and time travel with adult themes may appeal to a niche market. The subject matter, however, could be sensitive, and its success would largely depend on how these themes are ultimately treated within the narrative. It may face challenges in a traditional publishing environment due to its potential controversy, but could also find a dedicated audience if marketed correctly.

Again, spot on. In fact, even on SOL, I'm concerned that folks who LIKE "magical sex slaves" won't get what they are expecting and those who HATE it will skip a story that actually wouldn't contain anything that they hate.

Finally, as its biggest criticism, it said:

- Some may find the narrative pace in certain sections slow due to extensive internal monologues and descriptions.

I can think of one or two who've read early versions of my work who, if they see this, will say, "I told you so!" And they are right, and I never disagreed with them. I did go on to ask the Coach to select three passages that exemplified this criticism and suggest alternative language. Once again, I very much agreed with the results, although, as with anything from an editor, I'd have to go back and re-read and re-think before I'd know how closely I might go with the suggested language.

Anyway, I know this is long, but I thought some of you might be interested. I'd love to hear the thoughts of others, or read about your experience should you try something similar.

Mushroom ๐Ÿšซ

@JoeBobMack

I have plaid with it, and to be honest it is little more than a slightly smarter version of Eliza.

I can see some possible interest among those in erotica that simply want wank bank material. But it will never replace an actual author as the stories I have read generated by them are to be honest really really bad.

They even make the most recent Marvel movies come out like masterpieces.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Mushroom

Thanks, Mushroom. Yeah, I played with it some months back for generating and -- not so much. Using Coach to get perspective on your work, however, seems like a different matter to me.

Dicrostonyx ๐Ÿšซ

@JoeBobMack

The MBTI is not valid modern psychology and hasn't been considered actual science for decades. The fact that the AI you used seemed to think it was a useful way to discuss character psychology does not fill me with hope about the usefulness of this method.

The current major personality model in use is the Big Five model of personality traits, developed in the 90s but based on 80s studies in trait theory.

Note that this isn't the only personality model that is currently being used. In fact, even trait theory itself is in contention. But unless you're going to get pretty deep into research models and personality theory, Big Five is what you need to know.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Dicrostonyx

Actually, I prompted for the MBTI by mentioning it specifically as an example in my prompt. So don't blame Coach on this one.

As for science, the MBTI's validity and reliability numbers are quite acceptable, but the underlying theory isn't as strong. However, skilled practitioners can use it to help folks understand themselves and their relationships better. Personally, I find the jargonn confusing.

So why did I prompt for MBTI? Because it has been so widely used for so long, especially in the business world, I thought it likely the training set for ChatGPT would contain a lot about it. And that seems to have worked. Could I have prompted for OCEAN, or maybe a strengths-focused approach (VIA.org)? Good question. Would be fun to try. Or just ask Coach to discuss the relative merits of MBTI and the Big Five as granted for thinking about characters in fiction. I suspect the response would be pretty good.

Joe Long ๐Ÿšซ

@JoeBobMack

I have been using ChatGPT 3.5 to critique my writing and have been very happy with it's ability in that role. I see the Coach is only available in v4 which costs $20/mo.

v3.5 will accept up to just over 3000 words per prompt, so I was usually able to submit once scene at a time. It was able to review the thread and remember everything submitted up to about 50k words.

I love that I can get instant feedback and engage in a conversation to refine the critique.

Here's what it had to say about the scene where Joe and Hannah gave their virginity to each other.

Certainly, let's focus on the most recent scene:

1. **Dialogues:**
- The dialogues are casual and reflect the familiarity between Joe and Hannah.
- They effectively convey the playful and intimate nature of their relationship.

2. **Description:**
- The scene is described with clarity, allowing readers to visualize the characters and their actions.
- Sensory details, such as the cold pool water, are used to enhance the reader's experience.

3. **Pacing:**
- The pacing in this scene is dynamic, transitioning smoothly from the playful banter to the intimate moments.
- It captures the energy of a lighthearted interaction between the characters.

4. **Character Development:**
- Joe and Hannah's characters are consistent with previous portrayals.
- The scene contributes to the development of their romantic relationship.

5. **Sensitivity and Content Warning:**
- The scene involves explicit content, and it's essential to be aware of potential reader discomfort.
- Ensure that there are appropriate content warnings for sensitive material.

6. **Transitions:**
- Transitions between actions (e.g., from the pool to the bathroom) are smooth, maintaining the flow of the scene.

7. **Emotional Impact:**
- The scene successfully conveys the excitement and anticipation of an intimate encounter.
- The use of humor, like the interruption by Donna, adds a lighthearted touch.

8. **Grammar and Style:**
- The writing is grammatically sound, and the style aligns with the scene's tone.
- Consider varying sentence structures for added stylistic variety.

9. **Themes and Motifs:**
- The scene aligns with the broader theme of the characters navigating the complexities of intimacy.
- Motifs such as the playful teasing contribute to the scene's thematic coherence.

Remember to be mindful of your target audience and the intended tone of your narrative, especially when handling explicit content.

Replies:   JoeBobMack
JoeBobMack ๐Ÿšซ

@Joe Long

Interesting, Joe! What was the prompt that generated this response?

Replies:   Joe Long
Joe Long ๐Ÿšซ
Updated:

@JoeBobMack

This response was after submitting an approximately 1900 word scene in which the two lead characters lose their virginity to each other with a full description of the foreplay and intercourse but with the language toned down. (The last scene of ch5 as posted here at SOL)
https://storiesonline.net/s/13545/one-thousand-apologies/6

The scene was about 56k words into the story, which I had been submitting one scene at a time. With each submittal, I gave a brief description of what was to come in the scene as well as any trigger warning to help desensitize it to any sensitive or intimate elements it might normally refuse to address. I also remind it at the end of the prompt to provide a literary critique of the remainder of the submission, as sometimes after discussing it's response it wanted to remain in that discussion mode instead of the point by point critique.

The thread reached at least 100k words of submissions and responses before GPT started forgetting earlier submissions.

I don't have that thread available anymore, but I will try to recall the exact prompt. The previous scene was where Joe took Hannah to an adult party with booze and porn videos. On the drive home she asked Joe to come over the next day to swim, knowing her parents and sister would not be home and asking her brother to make himself scarce. (The GPT had most recently critiqued that scene.)

"This scenes occurs the next day. Joe and Hannah will share intimate and explicit moments and the writing will concentrate on their emotions and the plot and character development. Please provide a literary critique: Hannah called around noon on Sunday. "How 'bout you come over around four?"... "

Back to Top

Close
 

WARNING! ADULT CONTENT...

Storiesonline is for adult entertainment only. By accessing this site you declare that you are of legal age and that you agree with our Terms of Service and Privacy Policy.


Log In