There's been a lot of talk on Hive recently about AI. My most recent post on the matter was mostly in response to the recent Hive Watchers episode about if it's possible to detect AI. I'd would encourage you to go read it, but if you don't, basically my answer is no, it's not possible. Even if it is slightly possible now, as ChatGPT improves and as we get closer to true AI, no we will not be able to detect it. This is for a number of reasons. Again, I encourage you to go read my post about it if you are interested.
Anyways in this article I want to cover what I see as the three main uses of ChatGPT in regards to writing. Now there there's probably many many uses but I'm I see these as basically falling into three categories. These may actually be obvious, in which case I am rambling on mostly just to practice my writing. But if you want to humour me, read on.
These categories are:
- the research assistant
- the editor
- the ghost writer
I want to talk about these three categories, and I want to talk about if each category is an acceptable use when it comes to writing on Hive or maybe published writing in general. Does that sound good? Well, let's dive in.
the research assistant
Image by mohamed_hassan from Pixabay
This category is basically Google on steroids or a grad student who will help you with research. That is to say the analogy I'm making here is when you're a professor and you're writing a paper—because most universities require professors to publish x number of papers per year—you don't really have time to research every single small bit of data. If you're writing a history paper, for example, you don't really have time to research what is this person's exact birthday, when did he die, so on and so forth. So these small things you will give grad students say, "Hey look this up for me," and they will run and get the information, they will bring it back to you, you can incorporate that in your writing.
Pretty simple, right? This can range from very basic questions—for example, I might say, "What is the start and the end date to the Heian period?"—to more advanced requests for information about other studies on the topic. At this more advanced end I might say, for example, "What is the name of the theory that proposes history tends to repeat in cycles of 80 years or roughly every four generations? Give me both the name of the theory and a list of the major papers written on this topic in the past 50 years?" And then of course ChatGPT will return with the name of this theory and with these papers that you can then ask further questions such as asking it to summarize one of the papers.
I recently used ChatGPT in this research assistant role. If you read my Haiku post a few days ago, I mentioned a TV show that I used to watch when I was a teenager and I gave a few details about the TV show. When I was writing that article I did not remember the name of that TV show. I didn't remember much about it except that this main character could pause time. So I I went to ChatGPT and I told it... well here I'll show you.
This is an app for my iPhone that uses the ChatGPT API
Now you see my question gave just the barest details. But ChatGPT was able to give me an answer and an accurate answer. I took this answer and I went to Wikipedia to confirm it and incorporated that info into that post that I wrote.
I think probably most of us would agree that this use of ChatGPT is not only quite useful, but an acceptable usage. It is simply using ChatGPT as a source of information, ideally a source that we then confirm. We are still doing all writing at this stage.
the editor
Image by Lorenzo Cafaro from Pixabay
As anyone who has had any published articles in magazines, for example, knows, a good editor is a must and will do a few things. First, a good editor will check for spelling and grammar mistakes and they will then fix those mistakes. We do have spell and grammar check now built into the computer, but it's still not exactly uncommon for us to miss something here when we are doing it by ourselves.
Second, a good editor will check for the house style. For example, if the magazine uses AP style or Chicago style or maybe they have their own in-house style that's a combination various style guides. They will then check your writing against the style guide they use and will update your writing for this particular style. Oxford commas or no, double quotes or single quotes, em or en dash, if it's the New Yorker, they will take away hyphens and add those strange diaeresis marks (I make fun here; I actually love the New Yorker style), and so on and so forth
Lastly, a good editor might rewrite some of your writing. Usually nothing dramatic, but editors might substitute words, overused idioms, or rewrite a line or two.
I think ChatGPT can do all three these for us. Not necessarily well, mind, but it can do them. We can request ChatGPT "please correct my spelling" "please correct my punctuation" "please correct my grammar" and it will do so. We can ask ChatGPT "please update my writing to the 17th edition Chicago style" and it will do a pretty good job of that. At the most drastic end we could ask it to suggest improvements to our writing and then compare each edit and accept or deny it.
I think many people on Hive are already using ChatGPT as an editor. I've seen many comments on posts saying they used it to correct their grammar or improve their writing. I've even seen some foreign users of Hive write a disclaimer saying they used ChatGPT to translate their post into natural sounding English. Kind of like Deepl with an editor twist for that last one.
This category is where things get a little thorny. Is this acceptable? Is it not acceptable?
I would say.... maybe. Most of the time it should be acceptable, but it does depend. Simple corrections are probably ok, but if we ask ChatGPT to rewrite and suggest improvements, to what degree are the improvements? Did ChatGPT basically rewrite the entire article or did it just change a sentence here and there? Depending on this answer, it might be acceptable, it might not be acceptable.
Ghost Writer
Image by Andrew Martin from Pixabay
This is where we are having ChatGPT basically write the entire damn thing for us. Just as you might hire a Ghost Writer to write a book for you, at this stage we are having ChatGPT write for us.
The whole ghost writing business is a strange one. Famous or rich people who can't write but want to do a book hire a ghost writer to do it for them, then they pretend they wrote it. Well, some pretend, some admit it. It's a weird business. Some people attack it as unethical, but many people who are ghost writers defend it. I suppose they must make decent money from it so they have an interest in defending it. (On the other side, there are many ghost writers who are against the practice, like @adamada as she wrote here)
I think we can say it's a controversial business, anyway. My image—and this is written from the perspective of never actually having used this business before—is that in many cases (but certainly not all) the "author" does supply things to the ghost writer, such as many interviews in the case of an autobiography or story ideas in the case of a fiction book, so at least there is some contribution from the "author" and in a sense we can say they did have a hand in making the book.
So enter using ChatGPT as a ghostwriter. I don't think ChatGPT is advanced enough to take a lot of input (yet), such as the "authors" may give to real ghostwriters, making this case more one of ChatGPT writing the entire thing by itself. And as @tarazkp has pointed out often ChatGPT's writing is not so much original as badly recombining writing from all of its sources. If you want some interesting thoughts on the matter, Norm Chomsky talked in this excellent article about why this is among other things. (Asking for "reader mode" in your browser will get you past the paywall).
This usage here of ChatGPT I am completely against and I think most people on Hive probably are as well. Perhaps this usage doesn't deserve a downvote, but it doesn't deserve any upvotes either, especially if the poster tries to hide the fact that they used ChatGPT.
Acceptable or Not
If you can think of other uses of ChatGPT in regards to writing let me know, but every use I think of falls into one of those.
To sum up my thoughts, I think using it as a research assistant is perfectly fine. Depending on the complexity of the research, one should probably double-check the info given, but regardless I think this is ok. Using it as an editor is probably mostly ok, but it can be case-by-case and the poster may want to mention it in their post. Using it as a ghost writer on the other hand is unacceptable. I'm openminded and willing to consider acceptable uses of ChatGPT in this capacity so argue your case in the comments, but until I see a compelling reason here I am against this usage.
So what to Do About it
Ok ok, I know I promised we'd get to my ideas on this last time but again this post is getting a little long, so let's save that for the third part.
Basically I think because AI detection is impossible or soon will be (as I detailed in that first post) then we can't automate this (as HiveWatchers wants to do) and have to fall back into judging a post on quality and the author's reputation.
Anyway, we'll get into that next time. Thanks for reading this far and leave me your thoughts in the comments.
(title graphic made by me in Photoshop from this image by Alexandra_Koch from Pixabay)
❦
 |
David LaSpina is an American photographer and translator lost in Japan, trying to capture the beauty of this country one photo at a time and searching for the perfect haiku. |