Just What Is “AI”?

To talk about contemporary (unclassified) AI, we need to recognize that it’s not what you would think by the term “artificial intelligence.” What the billion-dollar companies are selling us is actually machine learning (i.e., studying data to find patterns) matched with Big Data. In other words, although the code for this is extremely complicated, the concept of the approach is simple. 

A truly massive amount of data is scanned and both qualified and quantified according to programmers’ parameters, and when asked to do a task, the AI algorithm (i.e., a step-by-step process to achieve a task) goes through that data and figures out what you probably want for that task.

Why and When the “Most Common” Works

Success breeds success, and we tend to promote accomplishments over failures. AI is great when taking the path most followed is productive. So, let’s say AI looked at 100,000,000 instances of workers’ “successful” injuries and found the most common, as well as their most common causes and the most common means of preventing them afterward. The resulting report would most likely be the foundation for extremely helpful recommendations of workplace safety protocols.

AI Has Offered Incredible Things

Among the incredible things AI has done, we have to applaud its work in helping medical diagnoses and treatment, working out mathematical proofs, and increasing productivity. I’m happy about the idea of AI’s helping air traffic controllers, structural engineers, and the space program. Honestly, the possibilities are endless.

AI Sucks for Other Things

For reasons I don’t really understand (I’m dumb.), AI companies, instead of putting the spotlight on the things AI does well, have marketed the supposedly wonderful achievements of AI in areas where it actually misses the mark, including, if not especially, creative content generation.

In short, instead of highlighting how AI can go through an avalanche of data to highlight useful patterns in usage, failure, and recovery, AI companies have been ramping up their basic spell-check and paint programs to pretend they can evaluate quality, not just consistency, and acting like taking 100,000,000 poems/images and putting them into a blender means AI can “write” a poem or “draw” an image.

Of course, what AI offers may seem new to you, but it is only unfamiliar. AI has to have gotten it from somewhere else, even if it hides that from you by swapping out words, like a high school sophomore copying their friend’s old paper with a few rearranged sentences and substituting “bubbly” with “effervescent.” 

AI Slop

This emphasis on what AI does poorly instead of what it does well is why we have the term “AI slop.” This means “new” poetry is made from a blend of the most common quotes from Shakespeare and the tritest of clichés, and AI-created images are made from the most uniformly followed drawing principles and often form logos that looks like buttholes.

Why I Care

I admit that I wouldn’t mind this so much, except I work as an editor, and it kills me (So dramatic!) to watch people settle for AI editing, which is awful, just because it’s free or cheap. A few years ago, when AI was the new-new thing, I made a personal evaluation of several AI platforms and found the “editing” often changed the meaning of what I was trying to say, didn’t fix complex grammar issues, and added stuff to “develop” my ideas that has nothing to do with what I was trying to say.

In the years since then, I have to say it’s only gotten worse. 

Because the model for “learning” is to assess the most common way to do things, AI editing is increasingly based on previous AI content, which is then used for determining whether something written or artistically created is “good.” Literally, AI editing is eating itself.

AI Cannot Recognize Originality

AI’s reliance on the most commonly used steps to make its algorithms means it will constantly seek the most familiar, the most over-used, and the most tired options. (Yes, programmers can muck about with the concept of “most common,” but this often leads to gibberish, not inspiration.) 

Selecting the most commonly used expressions is helpful (though hardly perfect) as long as you’re following basic and specific rules, like knowing how to spell “accommodate” or making sure “..” doesn’t show up at the end of a sentence.

It is the opposite of helpful when you want to go beyond a list of rules.

For a real-world example, I asked Quillbot to “Write a short poem about how love is like a flower.” This is what it gave me:

Love is like a flower bright,
Blooming softly in the light.
Petals warm with gentle care,
Fragrant whispers fill the air.

Roots that grow both deep and true,
Nurtured by the skies of blue.
Though the storms may come and go,
Love will blossom, ever so.

Hallmark doesn’t write poetry that hackneyed and wearisome. Seriously, ugh.

But, you may say, your prompt was really trite. OK, so I asked Quillbot to “Write a short poem about how love is like a Popsicle.” I got something even worse:

Love is like a Popsicle bright,
Sweet and cool in morning light.
Melting slowly, drip by drip,
A fleeting taste upon your lip.

Cold at first, then warmth will grow,
A fleeting joy that makes hearts glow.
Hold it close, don’t let it fall—
Love’s a treat, enjoyed by all.

We can see clearly that today’s AI not only cannot be original, it doesn’t recognize originality. It only recognizes “outliers,” and typically rejects them.

Content Creation and Editing

Finally, you may object that creative content creation isn’t the same as editing, and it’s not. But if AI cannot create original content itself, it cannot reward or promote it for others. So, while it can make suggestions, it can offer nothing new in brainstorming. It can make your sentences more native-speaker-sounding, but not more evocative. It can correct your spelling, but it cannot challenge your thinking to go in new directions. It can find your writing “free of mistakes,” but it cannot recognize when you’re onto something that deserves a closer look, it cannot applaud a plot twist, and it cannot dare you to go beyond your first “acceptable” attempt.

That’s what a human editor still does that AI cannot. In fact, the human brain enjoys new things, celebrates innovation, craves surprises, and embraces the unknown. AI would likely just mark them as anomalies to be avoided.