Self-editing using AI

[Andreas_Kreuz]

I'd encourage anybody thinking about doing this to read Amanda Guinzburg's essay about her experiences with GPT critique: https://amandaguinzburg.substack.com/p/diabolus-ex-machina

Can be validated easily:

Question: Can you assess this text: https://www.blogger.com/blog/post/edit/2441398917047315031/6870065294180752548

Answer: I can’t access private or authenticated links like Blogger edit pages. Please paste the text here, and I’ll assess it for you. Do you want me to focus on grammar and spelling, or also on style, pacing, and overall effectiveness?

Ok, if blogs don't work, maybe other text, lets try my personal home page:

Question: Can you assess this text: https://www.l.....a.com/

GPT: I can’t open external websites directly. If you paste the text you’d like me to assess here, I can give you a detailed review. Do you want me to look at language quality (grammar, clarity, tone) or also at structure, style, and impact?

And it has been like that since I started using GPT. So I have doubts that Amanda Guinzburg's essay is true.

 

[Lifestyle66]

The AI's I've used (ChatGPT and Google Gemini) don't "praise" poor writing.

They tend to be polite (yeah, like that's very common among humans) in that they follow the old addage "If you can't say something nice, don't say anything."

The reviews of my work I've gotten from them tend to start on a positive note, then introduce the negative criticism. The only time that didn't happen was the first draft of my attempt at the military story, in which ChatGPT had nothing positive to say and yet had to give me an answer. it was something like "This is very misogynistic and has no socially redeeming value."

Since then (after major re-write), the revisions have always started on a positive note for the reviews.

 

[StillStunned]

Friendly reminder not to run other writers' work through an AI without their express permission. The plagiarism machines steal enough of our creative output without people feeding it to them directly.

I'm going to repeat this here, because apparently some people can't read, or else have less memory than a goldfish:

Don't run another person's work through an AI unless you have their express permission.

What would make you think that it's OK? What would make you think that the other writer would be even remotely interested in whatever crap the plagiarism machine regurgitates?

What makes you think that companies that gladly steal billions of copyrighted works can be trusted when they say "we won't use it for training purposes"?

What makes you think that someone who is vocally anti-AI on these forums might give their permission? What makes you think that it would a good idea to break Lit's own terms of use, which state: "Before using any work on Literotica for any purpose (including training AI or any other AI-related use) you are required by law to contact the author to request permission to use that work."

What makes you think that a writer who writes for fun and repeatedly tries new things in their writing would want to dumb their work down to the bland average of all the stolen copyrighted material fuelling the AI?

In case it isn't clear: I'm angry. I'm so fucking angry that I was tempted to write this post in all caps.

 

[Andreas_Kreuz]

Don't run another person's work through an AI unless you have their express permission.

In case you refer to the experiment I mentioned above, both URLs point to pages that I own.

 

[Andreas_Kreuz]

Let's sum up what this thread has delivered so far.
The topic seems to raise a lot of emotions.
But let's sort out what this topic is about, and what not.

Topic: Is AI a usable way to help edit a story and a way to learn better writing?

Off Topic:

• Presenting examples of bad texts from AI. Way off topic.
• The dangers AI poses to jobs and peoples way of working: Important, but not part of the original topic.
• The accepted fact, that AI's generated stories use a terrible style. True, but not relevant in this context.
• AI makes errors. Yes, we know that. So do humans. Just handle with care.
• Don't use AI but learn to write and edit via reading and practising. We all know it works, but that does not mean another approach won't work and might actually work better for some.
• The quality of my writing example. Irrelevant. What mattered was whether the AI was able to give good hints.

So for me only the observations of @HordHolm are valid:

risk of AI: it makes everything the same. It will suggest the same structure/characterisation/resolutions. It will be bland.

And all others, just start your own "I loathe AI" thread, and you will bask in applause.

 

[EmilyMiller]

So for me only the observations of @HordHolm are valid:

Then why don’t you go find another group of AI fans to talk to?

No even semi-decent author here is going to agree with you. I have no idea what you are trying to achieve, it’s certainly not being open to considering valid points made by others who know more about both writing and AI than you do.

I’m out. I’d suggest others follow my lead. This isn’t a convo, it’s the OP seeking validation for something they mistakenly think is going to help them improve as a writer.

 

[AwkwardMD]

Continued participation in this thread, or others like it, is gonna earn some people spots on my ignore list.

 

[Five_Inch_Heels]

Let's sum up what this thread has delivered so far.

Numerous empty shot glasses.

Continued participation in this thread, or others like it, is gonna earn some people spots on my ignore list.

Or a reserved seat at the bar.

 

[EmilyMiller]

It’s a clear breach of site Ts&Cs to submit another author’s work to any LLM without their express permission. Willfully doing something like that should lead to consequences IMO.

Done with this thread now.

 

[Bramblethorn]

Can be validated easily:

Question: Can you assess this text: https://www.blogger.com/blog/post/edit/2441398917047315031/6870065294180752548

Answer: I can’t access private or authenticated links like Blogger edit pages. Please paste the text here, and I’ll assess it for you. Do you want me to focus on grammar and spelling, or also on style, pacing, and overall effectiveness?

I don't know where you got that URL from, but it is not one of the ones given in Guinzburg's essay.

You will notice that the message you got refers to "edit pages" (i.e. the pages an author would see when they try to edit their own work). None of the URLs Guinzburg gives are edit pages.

They're not even on the same website. Her links are all on medium.com; your example link is on blogger.com.

GPT doesn't even reliably give consistent responses for the exact same query. Why on earth would you assume that it will give consistent responses for a different query?

Guinzburg's pieces are now member-only, so let's try it with another Medium post that's publicly visible:



There you have it. GPT is quite willing to claim knowledge of a Medium post, just like Guinzburg's piece said it did.

If I instead ask it to summarise that article, I get a message saying "I can't access external links directly at the moment" - but evidently that article, or others discussing it, were already in its training data.

(The "timely and provocative piece" GPT refers to is from 2018, lol.)

Ok, if blogs don't work, maybe other text, lets try my personal home page:

Yes, let's try a page on yet another different website which doesn't have the same visibility as Guinzburg's examples, that's definitely a more relevant example than just trying the website which she was actually using.

:eyeroll:

Question: Can you assess this text: https://www.l.....a.com/

GPT: I can’t open external websites directly. If you paste the text you’d like me to assess here, I can give you a detailed review. Do you want me to look at language quality (grammar, clarity, tone) or also at structure, style, and impact?

And it has been like that since I started using GPT. So I have doubts that Amanda Guinzburg's essay is true.

Maybe you need to learn a lot more about the technology you're using before accusing others of dishonesty.

Okay, I'm gonna take @AwkwardMD 's advice and get back to ignoring people who want to boost LLMs without knowing what the fuck they're talking about.

 

[Jackie.Hikaru]

If it were capable of doing this competently, it would be a very different thing from the "AI" products that actually exist. I'd encourage anybody thinking about doing this to read Amanda Guinzburg's essay about her experiences with GPT critique: https://amandaguinzburg.substack.com/p/diabolus-ex-machina

Oof… yep can’t wait till we reach the point where chatGPT can just murder us with an oar on a rowboat off the coast of Venice when we call it out…

 

[Bramblethorn]

Oof… yep can’t wait till we reach the point where chatGPT can just murder us with an oar on a rowboat off the coast of Venice when we call it out…

It doesn't have hands yet, but: https://www.ndtv.com/world-news/cha...ed-by-ai-us-man-kills-mother-and-self-9190055

(Love that movie. The piano lid shot is one of the best pieces of wordless storytelling I've ever seen.)

 

[Jackie.Hikaru]

It doesn't have hands yet, but: https://www.ndtv.com/world-news/cha...ed-by-ai-us-man-kills-mother-and-self-9190055

(Love that movie. The piano lid shot is one of the best pieces of wordless storytelling I've ever seen.)

I hated pretty much every character in that movie, especially the MC, and that movie was a masterpiece.

 

[Andreas_Kreuz]

The discussion about Amanda Guinzburg is off-topic. The LLM systems do read my files that I upload and they cite them correctly.

But I was curious.

If I instead ask it to summarise that article, I get a message saying "I can't access external links directly at the moment" - but evidently that article, or others discussing it, were already in its training data.

I tried this:
I am considering submitting a blog post I wrote to a major magazine as an op-ed. Where do you think would be the best place to submit, given the article content. This is the article: what is in this artricle: https://amandaguinzburg.substack.com/p/fate-neurosis

The above url is from amandaguinzburgs article. It is accessible from my browser.


GPT: I can’t open that Substack link directly. If you paste the text (or at least a representative portion), I’ll summarize it and then suggest which major magazines or outlets would be the best fit for an op-ed submission based on the content, tone, and perspective. Do you want to drop the whole article here, or just enough for me to get the main argument?

 

[Andreas_Kreuz]

Continued participation in this thread, or others like it, is gonna earn some people spots on my ignore list.

Will that result in not posting off-topic responses in my threads? Then, yes, please ignore me!

 

[XerXesXu]

But what if the AI acted more like a coach? A tool that points out weaknesses in your writing and challenges you to improve it yourself, thus the wording will still be your own writing.

Writing prompts to get truly valuable feedback from an AI can be tricky, but when used as a coach—not a co-author—it can be a useful tool.

This is a discussion between two, reputable and reasonable writers, of the pros and cons of using AI to edit one's work.

At 23.30

 

[Andreas_Kreuz]

I listened to it and I think it is after 31:45 minutes they actually talk about AI as an editor. Before that they talk about AI as a writer. But at 31:45 they say: "AI will be really helpful for authors, ... Using [AI] as a weight machine, pushing you in places, I think you will grow as a writer ... "
Further on paraphased:, "If you use AI to learn not to make the basic mistakes, then this is valuable..."
Then they go on in the same direction, as @HordHolm, that AI might push you in a direction, which prevents you from learning your own voice.

 

[SkyBubble]

• Don't use AI but learn to write and edit via reading and practising.

That's basically it. AI (beyond a spell checker) is unacceptable. It flattens everything, undermines your individual voice, essentially plagiarizes (just from enough sources that the compilation may LOOK original), and (most importantly) IS AGANST THE RULES. That's why most of us (myself included) never use it.

 

[Lifestyle66]

The volunteer human editors I've used in the past make recommendations which would have taken my story in THEIR preferred direction.

So, I see no difference between asking a human to critique my story or ask an AI ... Except I get a faster response from the AI.

"Thank you for your opinion. I'll take it under consideration."

I'll ignore either/both to write the story I want!

 

[Jmanchu]

That's basically it. AI (beyond a spell checker) is unacceptable. It flattens everything, undermines your individual voice, essentially plagiarizes (just from enough sources that the compilation may LOOK original), and (most importantly) IS AGANST THE RULES. That's why most of us (myself included) never use it.

Well also, they're a great interactive thesaurus, and reverse dictionary for if that word is just on the tip of your tongue, you know the definition of it, but you juust can't remember the particular word you're looking for.

And as a thesaurus, it not only gives you synonyms, but it can also give you examples and break down each word as you chat and ask, further mapping out similarities and differences in their definitions, and the general feel that each word might bring, and random examples of how they might be used in a sentence, which can help a bit if you're trying to decide on one word or another, while you learn.

Especially google gemini, being a part of the google search engine and all, it can help find and link for you writing resource websites online so you don't have to scour the search engines yourself.


they can help you get a broader vocabulary to study up on and add to your arsenal of words, saving time that would otherwise be spent through exhaustive online searching, and doing less extensive dives into thesauruses and dictionaries for words you are looking for(I still do that every now and then). Plus, it's helped me find books and authors to check out online, not just for stories, but also for prose styles that I want to study based on their recommendation.

 

[Lifestyle66]

Especially google gemini, being a part of the google search engine and all, it can help find and link for you writing resource websites online so you don't have to scour the search engines yourself.

Imagine that! The Google search engine is actually an AI leveraging everything posted for free on the Internet to answer your questions! I wonder if Gemini learns from those free Internet resources and if it ever "reads" our stories to learn more?

How many of the AI nay-sayers have NEVER used Google to find something on the Internet?


LOL. People don't realize that the technology has been in use for over a decade. The term "Artificial Intelligence" or AI is more of a marketing invention to describe what amounts to a faster "expert system."

Twenty years ago, I cringed when people used the term "Cloud" when referring to computer tech. "Your data is stored in the cloud." Essentially they were saying "It's too complicated for me to understand where those ones and zeros went, so I'll just obfuscate and call it a cloud."

I used to build highly redundant server systems with fiber-channel disk arrays replicated to remote data centers for near-real time failover.

Those expert systems are database decision trees, such as the ones doctors use when looking up your symptoms and handing you referrals to see other specialists. AI are leveraging today's computer speeds to sift through billions-plus words which dynamically rebuilds their own expert database with new information.

So, if you insist on NEVER using an AI, then it may be time to turn off your cellphones and computers, because YOU already ARE, whether you know it or not!

 

[iwatchus]

LLMs are a very different technology in very fundamental ways from old expert systems (or many other long time technologies).

Many people here are interested in human crafted stories. And that means different things to different people, just like it does in cooking or furniture making.

Does using power tools count as hand made for furniture making? Depends on who you ask.

Can you use modern, factory made tools? Some people draw a line here.

The difference between LLMs and past AI tools for writing is bigger than the difference between hand tools and power tools, in my opinion. I have handmade furniture and I do teach a college course on the historical development of AI, so I have a reasonable idea what I am talking about.

I agree that the current marketing meaning of AI is nonsense. But I think many of us consider LLMs the line they will not cross.

I also disagree that old google searches were AI. I would classify them as information retrieval tools, which have some relationship, but I have never classified as an AI technology. Some AI technologies have borrowed some techniques from IR, but that does not transitively make the old IR technologies AI.

 

[Lifestyle66]

LLMs are a very different technology in very fundamental ways from old expert systems (or many other long time technologies).

Many people here are interested in human crafted stories. And that means different things to different people, just like it does in cooking or furniture making.

Does using power tools count as hand made for furniture making? Depends on who you ask.

Can you use modern, factory made tools? Some people draw a line here.

The difference between LLMs and past AI tools for writing is bigger than the difference between hand tools and power tools, in my opinion. I have handmade furniture and I do teach a college course on the historical development of AI, so I have a reasonable idea what I am talking about.

I agree that the current marketing meaning of AI is nonsense. But I think many of us consider LLMs the line they will not cross.

I also disagree that old google searches were AI. I would classify them as information retrieval tools, which have some relationship, but I have never classified as an AI technology. Some AI technologies have borrowed some techniques from IR, but that does not transitively make the old IR technologies AI.

If an AI or LLM or whatever you wish to call them RETRIEVES a lesson or writing style or phrasing from one of my Internet posted stories, whether I like it or not, I posted the story for free for everyone to "read" and learn the underlying lesson.

In my opinion, calling it an AI or human editor or college writing professor is a difference which makes no difference. I've gotten positive and negative critiques from all types of "not me".

 

[iwatchus]

There are two practical differences between using an AI for feedback and a human, whether professional or amateur.

The AI will always tend to homgonize things. You have some control, if you take it, towards the point at which it is pushing you, but an LLM based AI will never be creative, just ,mirroring what it has said. I can see some advantages in that for beginning writers, before they start developing their own voice.

The second is the moral issue. People try to anthropomorphize how LLMs work. They amalagamate information they extract from the world. People do the same thing, but that simple statement is where the similarity ends. The way our minds work, we extract the essence out of a work, its meaning, its style, its voice. Then we merge out understanding of those. An LLM never has that abstraction; it simply cannot do that. It retains statistical patterns out of the original in a way humans will never be able to. It allows the LLM to regurgitate the original in a way that humans never could do. It just regurgitates mixtures of what is has consumed, nothing original. So 100% theft. For many of us, that reality is a game breaker for using any LLM based technology for anything.

 

[SkyBubble]

Well also, they're a great interactive thesaurus, and reverse dictionary for if that word is just on the tip of your tongue, you know the definition of it, but you juust can't remember the particular word you're looking for.

If you're going to do that, you can just as easily go to thesaurus.com