I have the text of my children’s picture book back from my editor. She described it as “really sweet” and “a joy” and made minimal suggestions. I’ve been over the text one last time and think it’s ready to go. Since I’ve never done this kind of book before, I’m flying blind, but I think the next step will be to find an illustrator and, after the illustrations are done, to find someone to handle formatting, which I can’t do myself this time. I have some people in mind, but it would be inappropriate to say who until I get commitments.
I’ve also been toying with Grammarly GO, which is the new “A.I.” plugged into the Grammarly grammar-checking app. Grammarly was already built on the same technology as these new “large language models,” so the introduction of this new A.I. assistant was probably relatively easy. Like the rest of Grammarly, it’s set up more for business emails than for creative writing, though as I’ve played with it, I’ve found it more impressive than I thought at first. It has a few stock questions you can ask it, and one is “Find my main point,” the result of which is the image at the top of the post. But after a little work, I got it to make the following suggestions, which, though brief and vague, indicate that the A.I. has mostly processed the story correctly:
It has correctly identified this as a work of fantasy fiction, and it has also correctly identified the protagonist and the villain. Obviously–and unsurprisingly–it can’t distinguish a children’s fairy tale from an adult novel, so its requests for more detail are irrelevant. Its second and third suggestions would only be reasonable if this were a different sort of work. The first suggestion, however, shows the A.I.’s limits: The opening of the story is, in fact, already dedicated to “who Anastasia is and what kind of person she is.” The software apparently couldn’t pick that up.
I wondered if it always gave these same suggestions, so I fed it a longer, more fleshed-out novel, specifically Rags and Muffin.
LOL. Ouch. So now it’s asking for less detail. “Clarify the narrative focus” may be legitimate, though not quite in the way Grammarly GO means it, since the presence of some plot lines unconnected to the main story is one criticism I’ve received from real human readers. But characters who “feel somewhat undeveloped” is definitely not a criticism I usually get.
In any case, the impression I get of Grammarly GO is the same one I get of ChatGPT: Its suggestions aren’t entirely bad, but they’re vague and so elementary that I can’t imagine them being useful to any but a beginning writer. For anyone else, they are at best good reminders.
I may be wrong, but I personally suspect the hullaballoo over “A.I” will prove to be a tempest in a teapot. The company that owns ChatGPT is hemorrhaging money, and these programs apparently degrade over time: A.I.s that were once whip-smart eventually lose their ability to perform even basic maths problems, and because they are really nothing more that advanced versions of word-prediction software, they often “hallucinate” information, presenting fiction as fact. Grammarly, too, is beginning to give more and more incorrect suggestions, especially comma splices.
Besides all that, the creators of these programs deliberately lobotomize them to make them politically correct. Many people have demonstrated this with ChatGPT, which will coyly dodge certain topics or even lie outright to avoid stating inconvenient facts. Grammarly appears to be undergoing a similar sort of deliberate retardation: Previously, it had “inclusive language” suggestions, and enough people complained that they added the ability to turn those off. Now the original “inclusive language” suggestions have been expanded to no less than sixteen different ungrammatical but politically correct settings that are on by default and have to be manually deactivated. The weirdest and stupidest of them is “Show Ukraine Support Message,” an utterly inappropriate setting for a grammar checker:
Once Current Thing has ended and we’re on to the next Current Thing, this option will presumably change to something else, like maybe “Express Fashionable Disapproval of Republican Presidential Nominee”–which will, of course, also be on by default.