NYC Midnight : Creative Writing & Screenwriting Homepage
Forum Home Forum Home > GENERAL DISCUSSION > Creative Writing Corner
  New Posts New Posts RSS Feed - ChatGPT
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

ChatGPT

 Post Reply Post Reply Page  <1 678910 19>
Author
Message
Pages View Drop Down
NYC Midnight Addict
NYC Midnight Addict
Avatar

Joined: 10 Oct 2021
Location: Edmonton
Status: Offline
Points: 744
Post Options Post Options   Thanks (0) Thanks(0)   Quote Pages Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 3:51pm
Originally posted by taaaylor taaaylor wrote:

It has been trained on this musician's lyrics without his consent. It's an insult to art AND it's outright theft. Even massive artists can't protect their IP. What hope do the rest of us have? Please stop using this bot. It doesn't even have the means to detect if copyrighted information is poured into it 

I read that Sudowrite and possibly ChatGPT have been skimming the biggest database of free stories out there: Archive of Our Own, or AO3. This is primarily a fanfiction website. People post their stories for free, so no copyright issues, but it's weird to know that a story you (fine....I) wrote is being used to train an AI on how to write a story.

Originally posted by copperdogma copperdogma wrote:

ChatGTP, because it has a sense of understanding, knew the answer: it's mostly tradition.


ChatGPT is confident - more so than Google. But confidence isn't truth. If you ask ChatGPT a question, it'll give an answer even if the answer is wrong.

Also, once you've had the AI write a story for you, re-prompt it with "write this like a human would write it". You may get better results.


Back to Top
copperdogma View Drop Down
NYC Midnight Groupie
NYC Midnight Groupie


Joined: 16 May 2021
Status: Offline
Points: 240
Post Options Post Options   Thanks (0) Thanks(0)   Quote copperdogma Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 3:57pm
Originally posted by Pages Pages wrote:

ChatGPT is confident - more so than Google. But confidence isn't truth. If you ask ChatGPT a question, it'll give an answer even if the answer is wrong.

Also, once you've had the AI write a story for you, re-prompt it with "write this like a human would write it". You may get better results.



I think that trust will be at the crux of it's acceptance and usefulness. Maybe each result could come with a confidence score, and clicking it would explain why it's confident (or not).

Transparency and communication usually build the most trust.

Back to Top
taaaylor View Drop Down
NYC Midnight Black Belt
NYC Midnight Black Belt
Avatar

Joined: 14 Jul 2018
Location: Idaho
Status: Offline
Points: 6946
Post Options Post Options   Thanks (3) Thanks(3)   Quote taaaylor Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 3:59pm
Originally posted by Pages Pages wrote:

Originally posted by taaaylor taaaylor wrote:

It has been trained on this musician's lyrics without his consent. It's an insult to art AND it's outright theft. Even massive artists can't protect their IP. What hope do the rest of us have? Please stop using this bot. It doesn't even have the means to detect if copyrighted information is poured into it 

I read that Sudowrite and possibly ChatGPT have been skimming the biggest database of free stories out there: Archive of Our Own, or AO3. This is primarily a fanfiction website. People post their stories for free, so no copyright issues, but it's weird to know that a story you (fine....I) wrote is being used to train an AI on how to write a story.


This fills me with a total and blind rage. I write online (reddit) and am at significant risk of being stolen from, too. I'm tempted to type in a keyword of my serial that only shows up in my serial, because it will be a dead giveaway if I've been plagiarized but I don't want to give those a-holes my phone number. To me it's especially soulless to steal from fanfic writers, given the potluck nature of the community and the legal gray area of asserting copyright. So f**ked. It rots the entire ecosystem of sharing stories with fellow fans. Thanks for mentioning this, as I'm out of my fanfic days and had no idea.
SSC R1
250 Micro R2

SSC 2021 winner
Back to Top
taaaylor View Drop Down
NYC Midnight Black Belt
NYC Midnight Black Belt
Avatar

Joined: 14 Jul 2018
Location: Idaho
Status: Offline
Points: 6946
Post Options Post Options   Thanks (2) Thanks(2)   Quote taaaylor Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 4:04pm
Originally posted by copperdogma copperdogma wrote:

Originally posted by taaaylor taaaylor wrote:

This is a really cool* example of ChatGPT being confidently dead wrong. Try using quotes around a word to preserve the spelling so Google doesn't return "self-raising" and "self-rising" as interchangeable spelling variations as they are different types of flour.

It's self-raising flour in the UK, and it's used because it contains a very specific ratio of baking soda to flour: https://topwithcinnamon.com/how-to-make-self-raising-flour/

Quote Self-raising flour (a.k.a. self-rising flour in the US) is a common ingredient in English baking recipes, typically used for scones, pancakes and Victoria sponge cakes. It is simply a pre-mixed combination of raising agents and flour so that you don’t need to add baking powder or bicarbonate of soda to your recipe

*by cool, I mean not even remotely cool. Don't trust the robot.


Oh sorry I must not have explained it very well. Here was the actual interaction:

ME: Why is self-rising flour more popular in britain?

CHATGPT: Self-rising flour is more popular in the United Kingdom, particularly in traditional British cooking, because it was historically used in recipes for quick breads and biscuits. These types of recipes do not require the precise measurements of baking powder and salt that other recipes may need. Additionally, self-rising flour was more readily available in the UK, and was therefore used more often in cooking. The convenience of self-rising flour also made it a popular choice for home cooks.


But you're right -- I'm not sure if that answer is actually correct. That's one of the things I'm fascinated with: How trustworthy is it and in what circumstances? That will drastically inform how widely its used and for what.

But that's still inaccurate. It's used for more precision, not less. It's not even why it was historically invented. It took me two google searches to discern that. Seriously, you should be more critical of what you read from a bot like that. It's still in training, and if you correct it, it will apologize for its mistake and adapt its statement to the new information you gave it.

Here, now we all can learn about self-raising flour lolol: https://cotswoldflour.com/blogs/news/what-is-self-raising-flour-and-when-should-you-use-it

Here's some advice: test it on something you actually know A LOT about. It's much easier to see the holes in its communication then.


Edited by taaaylor - 17 Jan 2023 at 4:05pm
SSC R1
250 Micro R2

SSC 2021 winner
Back to Top
Pages View Drop Down
NYC Midnight Addict
NYC Midnight Addict
Avatar

Joined: 10 Oct 2021
Location: Edmonton
Status: Offline
Points: 744
Post Options Post Options   Thanks (0) Thanks(0)   Quote Pages Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 4:20pm
Originally posted by taaaylor taaaylor wrote:

This fills me with a total and blind rage. I write online (reddit) and am at significant risk of being stolen from, too. I'm tempted to type in a keyword of my serial that only shows up in my serial, because it will be a dead giveaway if I've been plagiarized but I don't want to give those a-holes my phone number. To me it's especially soulless to steal from fanfic writers, given the potluck nature of the community and the legal gray area of asserting copyright. So f**ked. It rots the entire ecosystem of sharing stories with fellow fans. Thanks for mentioning this, as I'm out of my fanfic days and had no idea.

The bot shouldn't plagiarize your stories, but it will use the meat of them for creating a story, dialogue, and/or structure.

As an example, it knows how to add a twist. How does it know that? By understanding what a twist is and how it relates to the bulk of the text presented. It 'reads' the story and understands that clues can be dropped that might indicate a twist later. It's not great at doing this, it's still in training.

By 'meat' I mean: on AO3 there's a lot of stories about the Omegaverse. The AI has scraped all these stories and understands the basic concepts of them (Alphas are in charge, Omegas are subservient, Betas are neutral) and can plug these elements into a story.

To be abundantly clear here, I don't support AI-written stories, but I think it's too late to stop it. Now it's a matter of what we do with the information, how it's integrated into our society, and how painful that integration will be in the short-to-medium term.

I feel like the AI is taking away something intrinsic to humanity (creativity) and making a bid to eliminate humans' need to participate in this creativity. Machines were supposed to do the menial work for us so we'd have more time to create, instead AI is learning to be creative (by scrubbing art and writing sites) so humans can do the menial work (which corporations find cheaper than buying machines. To them, workers are expendable and cheap if there's no healthcare or salary expectations).

Some companies are also looking into how to watermark AI-written pieces. If this gets some footing, then it'll be easier for judges to pick out AI-written work in challenges like NYCM's.
Back to Top
taaaylor View Drop Down
NYC Midnight Black Belt
NYC Midnight Black Belt
Avatar

Joined: 14 Jul 2018
Location: Idaho
Status: Offline
Points: 6946
Post Options Post Options   Thanks (1) Thanks(1)   Quote taaaylor Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 4:39pm
Originally posted by Pages Pages wrote:

Originally posted by taaaylor taaaylor wrote:

This fills me with a total and blind rage. I write online (reddit) and am at significant risk of being stolen from, too. I'm tempted to type in a keyword of my serial that only shows up in my serial, because it will be a dead giveaway if I've been plagiarized but I don't want to give those a-holes my phone number. To me it's especially soulless to steal from fanfic writers, given the potluck nature of the community and the legal gray area of asserting copyright. So f**ked. It rots the entire ecosystem of sharing stories with fellow fans. Thanks for mentioning this, as I'm out of my fanfic days and had no idea.

The bot shouldn't plagiarize your stories, but it will use the meat of them for creating a story, dialogue, and/or structure.

As an example, it knows how to add a twist. How does it know that? By understanding what a twist is and how it relates to the bulk of the text presented. It 'reads' the story and understands that clues can be dropped that might indicate a twist later. It's not great at doing this, it's still in training.

By 'meat' I mean: on AO3 there's a lot of stories about the Omegaverse. The AI has scraped all these stories and understands the basic concepts of them (Alphas are in charge, Omegas are subservient, Betas are neutral) and can plug these elements into a story.

To be abundantly clear here, I don't support AI-written stories, but I think it's too late to stop it. Now it's a matter of what we do with the information, how it's integrated into our society, and how painful that integration will be in the short-to-medium term.

I feel like the AI is taking away something intrinsic to humanity (creativity) and making a bid to eliminate humans' need to participate in this creativity. Machines were supposed to do the menial work for us so we'd have more time to create, instead AI is learning to be creative (by scrubbing art and writing sites) so humans can do the menial work (which corporations find cheaper than buying machines. To them, workers are expendable and cheap if there's no healthcare or salary expectations).

Some companies are also looking into how to watermark AI-written pieces. If this gets some footing, then it'll be easier for judges to pick out AI-written work in challenges like NYCM's.

Yes, this is absolutely true and I totally agree. It's important clarification too, so thank you. By plagiarism, I mean copying syntax, imagery, language structure, thematic concepts, etc. as a function of its machine-learning. We will definitely need a different word for this kind of IP-infringement
SSC R1
250 Micro R2

SSC 2021 winner
Back to Top
Pages View Drop Down
NYC Midnight Addict
NYC Midnight Addict
Avatar

Joined: 10 Oct 2021
Location: Edmonton
Status: Offline
Points: 744
Post Options Post Options   Thanks (0) Thanks(0)   Quote Pages Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 4:57pm
Originally posted by taaaylor taaaylor wrote:

Yes, this is absolutely true and I totally agree. It's important clarification too, so thank you. By plagiarism, I mean copying syntax, imagery, language structure, thematic concepts, etc. as a function of its machine-learning. We will definitely need a different word for this kind of IP-infringement

Yeah, it's likely learning all those things, but they're harder to learn than size-colour-item (big red ball, not red big ball) or whether an adverb is needed (smiled happily vs smiled sadly). The AI now has trillions of parameters, which is astonishing, so I suspect it'll learn these things uncomfortably fast.

One would hope that as it learns it develops it's own style, just like humans do when we read novels when we aren't writing. When we sit down to write it's likely that the author's style will leak into ours, but the more we write the more we develop an individualized narrative style.


Back to Top
copperdogma View Drop Down
NYC Midnight Groupie
NYC Midnight Groupie


Joined: 16 May 2021
Status: Offline
Points: 240
Post Options Post Options   Thanks (0) Thanks(0)   Quote copperdogma Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 5:24pm
Originally posted by taaaylor taaaylor wrote:

Saw this on my reddit trawling today. I feel it's relevant: https://www.reddit.com/r/Music/comments/10e9639/this_song_sucks_nick_cave_responds_to_chatgpt/j4q5ge6/

And, to be honest, proof of the lack of moral ethic of people using ChatGPT. It has been trained on this musician's lyrics without his consent. It's an insult to art AND it's outright theft. Even massive artists can't protect their IP. What hope do the rest of us have? Please stop using this bot. It doesn't even have the means to detect if copyrighted information is poured into it 

Wow, thanks for sharing that. What a beautiful, heartfelt response he wrote. I'm in awe.

If there's a silver lining to ChatGPT, it's caused a bunch of artists to deeply reflect on what they create, why they create it, and what it means to them and the world.

I've been thinking about this a lot as a (wannabe) writer.
Back to Top
Suave View Drop Down
NYC Midnight Black Belt
NYC Midnight Black Belt
Avatar

Joined: 25 Jan 2015
Location: Thailand
Status: Offline
Points: 25028
Post Options Post Options   Thanks (0) Thanks(0)   Quote Suave Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 8:34pm

Watermarks only work if you copy and paste.
Now if there were a way of inserting certain words in a pattern throughout a story,
sort of like the US dollar has the Eurion Constellation that a photocopy machine will not copy....
Back to Top
copperdogma View Drop Down
NYC Midnight Groupie
NYC Midnight Groupie


Joined: 16 May 2021
Status: Offline
Points: 240
Post Options Post Options   Thanks (1) Thanks(1)   Quote copperdogma Quote  Post ReplyReply Direct Link To This Post Posted: 17 Jan 2023 at 10:51pm
Originally posted by Suave Suave wrote:


Watermarks only work if you copy and paste.
Now if there were a way of inserting certain words in a pattern throughout a story,
sort of like the US dollar has the Eurion Constellation that a photocopy machine will not copy....

We covered that earlier in the thread, but I'll paste it here:

https://theamericangenius.com/tech-news/chatgpt-watermark/#:~:text=You%20can't%20currently%20use,stage%20of%20ChatGPT%20feature%20rollouts.

TLDR: Cryptogenic watermarking. "The key to making AI content watermarking virtually undetectable is ensuring the distribution of words still has a unique and random appearance like regular AI-generated text."

That's interesting. If they do it consistently enough within the text, anything you don't rewrite fully by hand would still contain that distribution ("watermark"), so the more you leave unchanged the higher it would rate on something detecting generated text.
Back to Top
 Post Reply Post Reply Page  <1 678910 19>

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.05
Copyright ©2001-2022 Web Wiz Ltd.

This page was generated in 0.188 seconds.