I must have misunderstood you somehow.
Clarke thank you for describing your process and using AI to bounce and clarify ideas. I liked your very specific process of creating your venn diagram and how it clarified your own thoughts.
For me the mental sparring and cocreating with ChatGPT is like magic.
You have ideas and thought. AI helps you develop your ideas or even gives you more ideas, then you are providing better copy for your readers.
A comparison with the advent of Photography:
Baudelaire described photography as āartās most mortal enemyā and as āthat upstart art form, the natural and pitifully literal medium of expression for a self-congratulatory, materialist bourgeois classā [13]. Other reputed doom-laden predictions were that photography signified āthe end of artā (J.M.W. Turner); and that painting would become ādeadā (Delaroche) or āobsoleteā (Flaubert) [14].
I use a lot of AI transcription. AI does a good job of cleaning the transcript. I can also, change the form or the tone. If I need to summarize my notes for a client, I am not against AI generated text (edited). AI has been good at pulling out data, creating markdown tables and interrogating PDFs.
1 Like
I did see the diagram; I saved it for future reference (with attribution if I ever share it with others 
). My only concern is the need to involve the FBI. 
With the caveat that Iām not a moral relativist (a position within meta-ethics impossible to hold consistently and coherently), it is not my place to tell you what is ethical in your approach to writing. To do so would require far more knowledge of the exact process than I have.
I re-read what you shared above. The genesis of my concern about having AI write a first draft stems from your initial comment:
I use AI to write my first draft (of my ideas). Sometimes thatās a quick chat about the ideas I want to share (theyāre my ideas), and sometimes I literally spend hours sparing with chatGPT trying to tease out my thoughts and make them more concrete and more valuable.
After more careful reading and more thought (always a good practice ), I better understand your approach. You are not publishing a heavily edited version of what AI wrote. You are publishing a heavily edited version of your ideas , which have been thoroughly vetted through dialectic dialogue with AI. Would that be an accurate description? If so, I would be far less concerned about the ethics of the process. My primary concern about AI and writing is publishing edited AI content as oneās own.
1 Like
I wonder if you āmisunderstoodā deliberately, Karl?
I donāt think you think AI is a āsomebodyā, so there must be something youāre not saying explicitly, and itās hard for me to follow your line of thinking.
Can you just spit it out?
I would genuinely love to know what you are thinking, even if I think differently.
I donāt want to put words in @karlnyhus mouth, but the AI is not you. Nor is it a person. It is āotherā.
Now there is room for some nuance here. But as these tools continue to improve that room will shrink.
If one were to give their executive assistant (EA) a topic and ask them to write a blog post, and then take that first draft and further instruct the EA to make a few changes, spelling and grammar, and make it in a more active voice. And then take the second draft and provide further instruction to add a touch of whimsey, would one be justified in considering the resultant final draft to be something they wrote? I would argue the EA is the author.
Replace āexecutive assistantā above with a ālarge language modelā. How does the conclusion change? Clearly the EA is intelligent where the LLM isnāt. But both are generating content. That is the point of generative AI now isnāt it.
So the question is, if the LLM is generating the content based on prompts I give it, can I claim authorship?
Before answering, consider not text, but graphics. If you give prompts to a generative AI to create an image, tweaking back and forth as with the blog post above, would you feel comfortable claiming to be the artist? Or do the same with one of these tools to generate music.
Clearly I can generate text (no LLM was used to generate this post), but I have no artistic or musical abilities. As Iād be hard pressed to claim to be the the artist or composer, I wonder if I can legitimately claim to be the author.
2 Likes
Thanks Steve. Thereās a lot of nuance here. And definitely a continuum.
Plagarism, which is, I think, Karlās concern, is definitely on one side of the continuum.
If one were to give their executive assistant (EA) a topic and ask them to write a blog post, and then take that first draft and further instruct the EA to make a few changes, spelling and grammar, and make it in a more active voice. And then take the second draft and provide further instruction to add a touch of whimsey, would one be justified in considering the resultant final draft to be something they wrote? I would argue the EA is the author.
Iād say it was a combined effort, though I doubt most people would really care who wrote it.
And, if you replaced EA with someone whose job title was Content Writer, itād probably feel a little different.
Iāve repeatedly asked my editors if theyād like me to credit them in my books, and they always decline. Theyāve told me that would feel unprofessional. They prefer testimonials and that I pay my invoices!
Arghghgh. Nuance! So annoying!
So the question is, if the LLM is generating the content based on prompts I give it, can I claim authorship?
Most of the time - at least, the way Iāve been using it - yes.
Some of the time, no. And, then, the write is most likely a big smelly pile of 
.
You could ask the same question about my spellchecker, I suppose. I canāt spell very well, and my grammar is pretty off.
BTW: I ask my assistant to prepare graphics for me. Sheās really good at it. She uses canva. Sometimes, when it feels appropriate, I add, āgraphics by NJā, othertimes it doesnāt feel right so I donāt.
Yes, thatās a good way of putting it.
Together we are cleverer.
(By which I mean you and I, though I couldāve also meant me and chat GTP).
1 Like
Exactly.
I do think of what we call AI today as a āsomebody.ā Artificial Intelligence for writing today is just a large language model comprised of the writing of thousands if not millions of actual intelligent people. How could anyone claim not to be using the work of others?
I donāt know of anyone whoās made that claim. Do you think I did?
Karl, Im grateful that you replied, but Iām still not understanding your logic. And Iām sorry about that. But thanks
Do you think that using ChatGPT is plagiarism - i.e. theft?
I asked you that question and you keep dancing around the answer.
Since Iām not using AI to write drafts for me, I donāt feel it necessary to pronounce judgement. But I will keep asking questions.
Iām not dancing! Iām confused.
If it helps: I donāt think itās plagiarism, anymore than I think using a spellchecker is plagiarism.
I would like to see authors and other creators (like me) compensated for using our words as the AIās fuel, though I donāt especially care about that.
Thank you.
ā¦
We got there Karl!
Why did you start talking about plagiarism?
It seemed a natural question to ask when someone announces they use AI to write drafts. And that after an unspecified number of rearrangements and word changes, and apparently without attribution, publishes the work as their own.
Oh. You said - I think - that you donāt have an opinion about this. But the way you write about makes it sound - to me - like you do.
Which makes me feel confused when I read your words.
It sounds - to me - like you do think itās plagiarism. Words are tricky though, so Iām still tbh not sure what you think. But Iām glad you took the time to elaborate.
Going back a bit: your process sounds like a form of working with a ghostwriter or co-writer. You are bringing the ideas, refining them with a partner, and doing developmental editing pre and post the first draft.
Writing and editing are nearly inseparable in early developmental editing. After the first draft, the line is more clear but developmental editors still have influenced the final wording.
Hopefully Iāve worded this neutrally enough to be resented by all.
4 Likes
This is an interesting discussion, and I think it hints at something that has bothered me for a while: itās clear that some people find that AI has big benefits for them as a writer; what isnāt so clear is what benefits that brings to the reader particularly in terms of trust.
On the Scrivener forum a week or two ago I proposed a simple thought experiment, which helped me think through the issues.
The basic idea is:
a) Any form of writing from now on must include a checklist detailing how AI was used in its production
b) All writers, being the moral paragons that we are, will complete the checklist with complete honesty.
The original post gave a list of the main ways in which AI is used in order of āintrusivenessā[1] : e.g.
- spellchecking / grammar,
- suggesting rewording to make characters/dialogue more realistic,
- helping with research,
- āconversationsā to suggest plot improvements or to construct the plot
- writing part / all of the book
- generating artwork
So, the author fills the checklist in with scrupulous honesty: how many ticks would it take for you as a potential reader not to buy the book / read the article?
Yes, of course, itās not going to happen ā Schrƶdinger didnāt put a cat in a box either ā but it did help me think through the question of which uses of AI I would be happy to reveal explicitly to readers that Iād used.
And if Iām not happy revealing the extent, then arenāt I breaking an implicit trust that should exist with the reader? Perhaps I shouldnāt be using those tools at all.
it was aimed at fiction, but could be adapted for non-fiction easily enough ā©ļø
1 Like
Itās interesting reading that. It was good of you to share.
I think itās good for writers to think about such things, but Iāve got a different suggestion:
I would like to make it mandatory that the AI tech companies include a modern version of MS Clippy, that pops up with comments like this:
āIt looks like youāre trying to write something that other people will read! Would you like me help you make it readable, engaging and useful?
We could start by converting your writing from passive voice to active voice. Iāll teach you why and how. And, if you want, I will show you before and after so you can learn.
And then Iāll help you reduce your word count by 15% and Iāll explain why and how. Your readers will love you because you save them time!
I know writers worry about stuff like chatGPT, but if chatgpt and its friends can teach loads of othet people to write better, that would be fantastic.
Good for everyone.
Iāve started asking ChatGPT and Claude to help me improve my writing and explain why and Iāve learnt so much.
I donāt take all of its advice but Iāve genuinely learnt a lot.
ā
I am helping a 57 year old dyslexic friend use chatgpt to transform poorly written words (which drain him, and cause him so much stress) into words that say the same thing but are easy to read. A life changer. Wonāt it be cool when everyone can do that?
Thanks! I do appreciate that there are benefits for writers from using AI and you outline them nicely.
But Iām specifically interested in the effects on readers. Given the checklist I suggested, is there a point where there are too many ticks for you not to read further?
For example, are any of the items on the checklist so fundamental to the role of the author that admitting to using AI undermines the faith of the reader in the basic competence of the author?
1 Like
Iāve taken a bit of time to think about this, and I originally concluded that I wouldnāt care about any of them, because, whatās important to me - as a reader - is that the content is helpful and well written.
But then I thought a bit more and I decided my answer was lazy thinking.
Iād want every author to use ALL of them - provided it helps them write BETTER BOOKS, and if it helps them write MORE BRILLIANT BOOKS.
Itās such hard work writing - physically and mentally - and so fianancially unrewarding for so many authors, why should we have to suffer?
On the other hand, if the technology makes it easier to produce more WORSE books ā¦ I hope the market soughts that out, and blames the lazy author, not the tools the good authors use.