Editing companies need to stop pretending their spelling and grammar checkers are “artificial intelligence.” The claims they are making regarding their AI software are patently misleading and often false. No AI editing can match the quality delivered by human editors with PhDs in their disciplines.
Considering the current marketing bonanza of slapping “AI” on everything from watches to shoes, it’s no surprise no one can agree on what AI actually is. According to the US Future of AI Act, a company can legally call their products AI if they “can learn from their experience and improve their performance.”
The keyword is “learn,” which means to qualify as AI, a program only has to take in data, assign tasks to that data, and then act on that data according to its programming. In current AI editing programs, this means examining lots of documents, figuring out how things are “normally” written, and applying that normality to a new document. Because the program grows more (and less) accurate in applying normality as it examines more documents, it is technically “learning” (which includes learning things that are wrong).
And that’s it, people. That’s all these programs do.
But you’d never know it from the promises editing companies make while touting their “lightning fast” and “super reliable” AI editing programs. Poorly understood and ridiculously overhyped, AI has become such a mighty and magical force I wouldn’t be surprised to see it featured on a box of cereal.
One company I won’t name so I don’t get sued goes so far as to claim that because their AI has analyzed billions of documents, it corrects grammar “intuitively.” A grammar checker has no intuition. But legislation controlling what companies can say about their AI products is dragging years behind the need, so I guess software designers can claim their AI will weep with joy at the greatness of its own abilities and not get into trouble.
Let’s go back into the annals of time, namely 1985, when Microsoft Word Version 2.0 was released with added features, including spellcheck and word count. It’s telling those two new abilities were paired up because they’re equally (un)impressive. Word 2.0 could take each clump of letters and count it as a word, and Word 2.0 could take clump of letters and compare it to a dictionary to make sure it matched up. Of course, it could only tell you that you’d spelled “there” correctly, not that you meant “their.”
Over the years, Word added a grammar check, and in so doing laid the groundwork for the AI editors of today. The difference, again, is that “learning” function. Word’s grammar check compares the work to a (necessarily limited) list of grammar rules. AI editors form and reform their own list of grammar rules (right or wrong) by examining documents.
That distinction may be legally acceptable, but this is so far from what we’ve been told “artificial intelligence” was going to be like since The Jetsons that it really seems software developers need to come up with a new name. I suggest “Super Clippy” or “Your Plastic Pal That’s Not Fun to Be With.”
Hee hee. Nerd humor.
In short, we’re being promised something superhuman for a product that can’t even walk a straight line, let alone fly.
And it’s not like the ineptitude of AI editing (let alone writing) is a secret. Media content creators keep getting fired for using chatbots without double-checking for mistakes, and when I did another round of research for this blog post, I repeatedly enjoyed the irony of reading about the expertise of AI editors with text that was full of errors. I can only assume they used an AI to edit it.
Here are some examples. Again, I won’t link them up so I stay out of trouble, but you’ll find your own errors easily enough if you look:
- Write confidently with AI powered editing. (“AI-powered” needs a hyphen.)
- Enable G***y wherever you type (This is a sentence, but there’s no period.)
- All-in-one writing assistant for desktop and browser. (This is not a sentence, but there is a period.)
- Creators use U***i to quickly turn keywords into full-length mind-blowing articles. (There should be a comma after “full-length.”)
- Write faster with AI suggestions — while remaining true to your unique style. (There is no reason for that em dash.)
However, the greatest irony in all this is how AI marketing is constantly promising that it will “make your work stand out” and “put you ahead of the competition” when, in fact, AIs ensure your work will be just like everyone else’s: bad grammar, lazy ideas, bad facts, and all. Copying other people’s idiocy is literally a function of AI’s programming.
And despite the inferiority of AI to, say, a high school senior with a grammar book, editing companies want people to shell out money. (At least free AI editing is worth what you pay for it.) I’ve noticed many sites now let you can pay a set amount to use a website’s AI program exclusively or pay more to have a human eyeball it afterward. Some let you pay less money for a “basic” AI service and more for a “premium” AI service that makes no guarantees about better accuracy (It can’t.) but will offer more suggestions, helpful or not.
One day, if we don’t all kill each other while we’re arguing over the proper use of AI, artificial intelligence is going to reach Jetsons-level abilities, but the future is not now, and it’s not going to be anytime soon. So far, AI is just a fancy term for an editing tool that only works when it’s used correctly. It’s not an editor, and companies should stop saying it is.
All this means I can say with confidence that real editors are all 100% human, like the PhDs at Edit911.com/Staff. In fact, Edit911.com uses no AI at all.