Over the last few months, WordPress has unveiled a new artificial intelligence (AI) tool designed to give feedback on written blog posts. Ultimately, my curiosity got the better of me and I decided to apply the AI feedback tool on a large enough sample size of my articles that included both older and newer works. Truthfully, I didn’t change anything featured in my articles; this was more of a I wonder what it’ll say type of affair. Before I knew it, I was on an AI bender.
Broadly speaking, I am curious as to what the AI was trained on. Much of the feedback that was generated offered suggestions such as including a summary as a concluding paragraph or to include sub-headers. This leads me to believe that a large part of the AI training material are essays from high school English class or BuzzFeed-style listicles (barf) rather than pieces written by authors with serious career ambitions. Given that, many of my articles were given the compliment of being “thought provoking” (such as my Dubbing Philosophers article) or “engaging” (like my article on being “The Man”). However, I wondered what was the benchmark that I was being compared to. If I’m right in my high school essays and BuzzFeed guess, and my articles were simply being compared to millions of students and uncreative drones, then it isn’t much of a compliment; out-witting a teenager or a BuzzFeed writer (a term I’ll use loosely with that god awful site) is not something to boast about. With that said, if the AI’s training base is fellow WordPress users, then that does lend more credibility to the feedback.
Many of my articles that were apolitical in nature and more linguistic focused, such as the In Critique of “Get Your Shit Together” article and the dancing monkey article insisted on calling me creative a large part of the time. Also, the bot kept giving me positive pieces of encouragement such as “Keep up the great work!” and other such mantras. I had the bot analyze my article of finishing my second book and it gave me a far-too-cheery response. The AI gave me little constructive feedback and instead was heavy on the congratulations and praise. It was as if the bot became my personal cheerleader instead of the critique-dispenser that it was supposedly designed to be. For these written works, its as if the bot’s main objective was to keep me writing as often as possible. If I was more of a conspiracy nut, I’d say that WordPress baked that positivity into the bot to encourage future renewal payments; after all, who doesn’t like a good dopamine hit?
Many 75 IQ pundits (eww…pundits) love to proclaim that AI has no bias, though that statement is just absurd. Don’t believe me? Well, The World Economic Forum agrees that bias is present in AI as well, and the WordPress bot is certainly no exception. Many of my more opinionated articles, such as How to Fix Gambling and Let The Kids Margin Trade, resulted in the bot giving me feedback along the lines of consider presenting a more balanced stance on the issue. This even happened in articles of mine that did feature my responses to potential objections. It seems as if the bot has a bias for neutrality, and a tendency to deliver more critical feedback on pieces written on a spicier topic versus a tame one. Another example of this is when I suggested that The Muppets make a skit in which Miss Piggy dates Andrew Tate, as the beginning of the feedback was “controversial” and “problematic”. My wish for news networks to drop all false pretenses of neutrality was a “provocative viewpoint”. Other articles of mine were critiqued as “having a strong Libertarian viewpoint” in the first sentence of the bot’s output (apparently that’s bad?). Essentially, if a librarian or English teacher would find the piece to be idealistically unhinged, then expect harsh judgment from the bot.
For some of my spicier articles, the feedback was skimpy on anything grammar or flow related, and instead was favored much more towards the bot becoming The Tone Police. This was readily apparent in my Legalize Home Defense Landmines article, which is undoubtedly one of my spiciest articles-and one that I absolutely stand behind. The feedback points were denoted in unusually present asterisks; the bot wanted to absolutely make sure that I noticed what it had to say. The bot focused its attention on my “disrespectful” and “dehumanizing” tone, as suggesting that I “consider the ethical implications” of my idea. My article on why I love radicalization yielded feedback from the bot that focused mostly on my “manipulative tone”. My article on why intermittent fasting is the new veganism ended up with the bot advising that I take a less “mocking and satirical tone”. Even my article on the ESPN Layoffs suggested that I “tone down the aggressive language”. Suffice it to say, the AI tool labeled me as a confrontational writer in many of my articles.
In fairness, much of the AI bot’s heaviest critiques align with in-person criticism that I’ve received. Those who have read my first book have called my tone and word choice aggressive. Additionally, human readers of some of my spicier ideas have critiqued my writings as well for being unbalanced. Simply put, I’m not calling the AI incorrect by any means. However, I have noticed that the AI tends to align towards the sensitivities of polite society; this makes sense, polite society is likely what the AI was trained on. Hence, I learned the hard way that the inherent biases of what an AI is trained on will inevitably be replicated by the AI. With all that said, read-aloud remains my favorite automated writing tool; read-aloud doesn’t judge.
Regardless of what feedback I received, cheerleading or condemnation, I became hooked on simply seeing what would come out of the bot. Admittedly, I started to dig deep into my past works to find the spiciest content that I’ve ever written simply just to see what the bot would say. It’s as if I was subconsciously hoping to frazzle or otherwise break the bot, regardless of how ridiculous that is. However, as fun as that was, I need to be cognizant of not letting the tail wag the dog as far as allowing AI to influence my future works. This AI bender has taught me that it doesn’t really matter what the AI was trained on, my brain working differently than yours is what separates me from other authors; my oddball stances are a feature and not a bug. Therefore, I don’t think I can use this tool, outside of its inherent entertainment value, to guide my writing.
Keep your AI, I’ll take my special sauce every time…
Home Page 