I wanted to talk about two things tonight. The more important thing is buried deeper down.
First, I mentioned a few nights ago that I’ve been taking my downtime to study the works of mega bestsellers and have been learning a lot. Someone asked me yesterday who I’ve been studying and some lessons learned.
First, I’ve been studying John Grisham. I like Grisham a lot and he’s easy to study. He lays it all out on the page for you to study if you know how to “see” craft techniques. He’s similar to James Patterson and Michael Crichton in that regard. Nora Roberts, for example, is a bit trickier to study because she does some amazingly advanced stuff. So does Dean Koontz.
Anyway, Grisham is great to study if you want to learn how to introduce characters. Not so much the “how”, but the mechanics of it. You can see clear differences in how he introduces main characters, supporting characters, minor characters, and even cameos. Very useful to internalize and master.
I’ve also been studying Brandon Sanderson. Not only is he one of the bestselling fantasy authors alive, he’s also one of the best practitioners of the craft in fantasy. Sanderson is famous for his magic systems (they’re great), but something that he also does really well are character tags.
(A character tag is a physical or personality trait the character has that the author repeats to make them more memorable.)
I’m rereading The Rithmatist, and the tags really jump out. They’re very easy to study in this book. You literally can’t miss them. By learning how Sanderson does it on a basic level, you can apply this to your fiction too and it will make your characters more engaging and memorable.
Anyway, that’s the first thing.
The second thing is that I mentioned that I was playing with ChatGPT-3. I don’t really need to say anything about ChatGPT-3 that hasn’t been said elsewhere. It’s an amazing tool whose best opportunities haven’t been realized yet.
Google ChatGPT-3 and you’ll see a ton of ideas on how to use it, mostly around content creation, questions to ask it, and so on. However, I was watching a video by Nick Thacker yesterday and he mentioned something that I haven’t seen anyone mention that is waaaaay more important: ChatGPT-3 can also edit text.
If I use the prompt: “Edit this text”, ChatGPT-3 will edit the text for clarity and typos. It will change the words most of the time (which I don’t like), but it will edit the text just as if you ran it through Grammarly or any other common spelling and grammar checker. Booo…
I played around with the prompt and changed it to “Edit this text for typos only”. Those three words made a gigantic difference. ChatGPT-3 will ONLY edit the text for bona fide spelling and grammar errors. It won’t change any of your words unless they are wrong. That’s the kind of editing you want.
So, that begs a couple questions. How good is it? Why wouldn’t you just use Grammarly or ProWritingAid instead?
I ran two tests.
In the first test, I took my last manuscript and pulled out the unedited sentences that my editor flagged for one reason or another—typos, grammar errors, etc. I ran all of those sentences through ChatGPT-3 with the prompt “Edit this text for typos only:” The test was to compare ChatGPT-3’s suggestions to the suggestions I received from my editor.
Sure enough, ChatGPT-3 caught approximately 60% of the errors. Keep in mind, this text was already sent through Word’s Editor, Grammarly Premium, ProWritingAid, and PerfectIt. These errors that ChatGPT-3 caught are all NEW and outside of the realm of what current tools can catch.
In the second test, I ran a 1500-word excerpt from my current WIP through ChatGPT-3 after running it through Word’s Editor, Grammarly, PWA, and PerfectIt. Sure enough, ChatGPT-3 caught 11 additional, bona fide errors.
Here’s what this means:
1. ChatGPT-3 can be another layer of defense you can use when self-editing.
2. ChatGPT-3 can help you create cleaner manuscripts before sending them to your editor.
3. When used with good self-editing and current tools like Word, Grammarly, PWA, and PerfectIt, ChatGPT-3 can free up your editor to look for more significant problems in your story rather than focusing on routine spelling and grammar errors.
It CANNOT replace an editor, and I strongly discourage people from taking that line of thinking. The value of an editor is far more than spelling and grammar errors, and ChatGPT-3 can’t catch character issues or plot holes, for example. But it’s another tool to help you catch the basic stuff. And right now, it’s free.
The only downside? It’s kind of a pain in the ass to use. You can only run about 500 words at a time through it, and it doesn’t give you tracked changes. In order to determine what it corrected, you have to use Microsoft Word’s Compare feature to enable tracked changes. It’s slightly tedious and time-consuming, especially for a full book, but maybe not so bad if you do it as you write instead of waiting until the end. But trust me, it works and it works well. It aced every test I threw at it, and trust me, I tried to trick it.
It seems to me that a writing app developer could find a way to integrate ChatGPT-3 into an existing writing app, so that they could use the OpenAI API to pass the text through ChatGPT-3 and then return the final, corrected text into the writing app as tracked changes. But, most writing apps don’t support tracked changes except for Papyrus Author. So, the best option would probably be an add-in to Microsoft Word that accomplishes this. I don’t know how easy or hard this would be to develop, and it may not be wise to do right now since ChatGPT-3 is just a preview. But I think it could work.
Anyway, that’s what I’ve been doing. I also made great progress on a story today and hope to finish it tonight or tomorrow.
