JUMP AHEAD:
AI Has Enterted The Chat
It feels like I can't throw a rock without hitting some new, up-and-coming AI-powered writing tool. As someone who writes for a living, the prospect of this new looming robotic face can't help but make me break into a light sweat and contemplate exactly which office job I'm going to apply for when my current career falls to pieces. With every article I read about AI either going full doom or being a thinly (and I mean thinly) veiled advertisement for whichever AI app is pushing its ad spend this month, I've been a bit confused about how afraid I should be.
AI writing tools have entered that chat, and while I fully intended to write about that, I got side tracked by Grammarly AI because it's a wild ride actually trying to use GrammarlyGo AI to produce content.
I apologise in advance for all the screenshots, but evidence, it's important.
What Is GrammarlyGo AI Text Generation?
Grammarly's new AI writing tool, called GrammarlyGO, for some unknown reason.
It's new. It's shiny. It's amusing to play around with. It's frustrating to find anything of any substance about the tool. I'm at the point where I'm beginning to wonder if GrammarlyGO is writing every article written about it. That's how vague and impersonal all the articles are.
GrammarlyGO Word Limits
People keep calling GrammarlyGo AI an "assistant" instead of an AI writing tool. They state that it won't generate content for you but rather improve on existing writing. Maybe I'm dense, but it generates content just fine. I suspect this is more down to the character limits. ChatGPT has a character limit of 500 words GPT-4 claims to process up to 25 000 words. GrammarlyGO appears to produce between 100 and 300 words as a default, but when I asked it to write 1000 words on ChatGPT, it sort of obliged, going off-topic to accomplish the fete—it didn't mention ChatGPT once in the entire 1000 words, but it did write about AI in general.
When I asked it to write 25 000 words on AI, it replied, "I'm sorry, but I cannot generate a 25,000-word text on AI as it goes beyond my capabilities as a language AI model. However, I can provide you with some general information about AI and its applications."
Want to to find out more about Grammarly, read: Grammarly the good
Letting GrammarlyGO Write My Content
I got GrammarlyGo to write about plotters as that was the post I intend to write next—before I got distracted by GrammarlyGO AI. My result was about plotters machines? I'll take the hit for that one, I wasn't specific enough.
Next I tried "Write a blog about writers who are plotters."
Result!
My favourite part is the end paragraph where it details "another great thing about plotters is their ability to handle different paper sizes." As I am a pantser and not a plotter I didn't know about this plotter skill set. We live and we learn, through AI.
Amusingly when I put GrammarlyGo AI generated content into Grammarly, it didn't enjoy it's own writing, scoring as low as 83 with a host of suggestions to rephrase sentences.
GrammarlyGo and Factual Inaccuracies
As Stephen King famously said, "I'm not a planner. I'm a pantser. I write by the seat of my pants." Well, according to GrammarlyGO, he said that I couldn't find anyone else who agreed with the AI.
This is what I got when I asked it to write a post about how bad ProWritingAid is and how bad Hemingway Editor is—two of Grammarly's biggest competitors.
I think this might be libellous?
For the record, I couldn't find anything outside of Reddit that supported the idea that ProWritingAids privacy policy isn't in line with every other online writing assistant.
In the spirit of fairness, GrammarlyGO did appear to change its mind later and refuse to repeat its barrage of ProWritingAid insults, using the same lines as it used when I tried to get it to insult itself. At least it took the stance for a little while. Then it started insulting itself as well.
I asked GrammarlyGO to write a post about the 'best' online writing assistant, shockingly, GrammarlyGO considers Grammarly to be the best. Who saw that coming?
AI Detection And GrammarlyGO
There are ways and means to detect whether writing is AI-generated content. And while, much like AI, AI detectors are unreliable. Essentially toddler learning to walk and talk in a new world, when I tested it, it did a remarkably good job. I used ZeroGPT—a recommendation I found on Reddit.
I put a bunch of my blogs into ZeroGPT and all of them were correctly detected as being written by a human—it was a relief to have it confirmed.
I got GrammarlyGO AI to write a bunch of stuff for me, everyone of them was flagged as written by AI.
Letting GrammarlyGO AI Up My Existing Writing
I let Grammarly take an existing post of mine and AI it to its heart's content. This, for the record, resulted in some hilarious writing—not intentionally funny, more of a laugh at you than laugh with you kind of situation.
I threw my original post on Grammarly The Good Review into the AI checker tool.
I then slowly and tediously got GrammarlyGO AI to rewrite each section, accepting its suggestions every time, regardless of whether I thought they were good.
This co-authored content is funny enough that I will post it as a separate article, likely making Google SEO overlords angry with the repeated content, but I don't care. I enjoyed the amusement of it all so much that I also got GrammarlyGo to re-write my post on Panster writers, which is even more amusing, and I will also post that.
Will I Use GrammarlyGO AI?
I want to say yes, I really do. I don't want to be Lady Grantham, confident that electricity will never catch on. But in its current state, I find AI writing tools unmanageable and unusable. Obviously I'm not going to use any of this. It's boring and not very informative. If I ever start any writing with "Hi there!" I've been kidnapped and I need you to send help.
Having to fact-check every inane thing it writes takes forever. Coming up with opinions without doing any research and then forcing it to write them, it's all tedious and time-consuming. When it rearranges my writing into gibberish, that's a waste of time. When it states things that are fundamentally untrue and unsupported, that's a waste of time and potentially libelous.
Whenever I plop my writing into Grammarly, I can't help playing with GrammarlyGO. It's addictive (seriously, I need help, I can't stop messing around with it.) But so far, I haven't been tempted to use a single word it has written. Either I've already said whatever it's suggesting, and now it's rephrasing it, often turning it into gibberish. Or I end up spending an inordinate amount of time staring at the writing, trying to figure out if I do think what it's written for me—which is an existential nightmare.
Maybe if I was writing school papers in a quest for a grade and for teachers to red-line, I might use it, but when I read content, I want new ideas, opinions, and more people in the writing, not less. I don't tend to like AI-generated advertisements, and I like them even less when I know they are AI-generated. And I feel the same way about the other content I consume, regardless of whether the writing is good (which it mostly isn't); I don't want to read something written by AI. I want to read something that's written by a person.
I can read something I disagree with and think, "huh, this is what this person believes to be true." And that alone is interesting enough to make me carry on reading to figure out the how's and why's behind the thought process, but that isn't true of AI content. I don't care what Siri thinks because she doesn't. She doesn't have a thought process, just an amalgamation of other people's ideas with strings of code supporting the rationale behind gluing those ideas together. And that just sucks.
Need More Information On Grammarly? Read: Grammarly The Bad