Empty Calories
So you don’t have to, I’ve been reading (actually listening to) a couple of hundred AI stories, mostly YouTube podcasts. I sampled sci-fi, dramas, true crime, romance, kinky romance, westerns, and, weirdly, kinky westerns.
The results? As you might surmise, early stories were deeply flawed, texts with plot holes so large they stretched the Bermuda Triangle into a quadrilateral. A few ’stories’ had no plot to speak of, just premise and exposition. “Jack loved Jill and Jill loved Jack and they lived in a shack on the hill out back where she took the pill and they got chill. The end.” Or some such.
Earliest AI attempts often confused gender pronouns, that is, traditional pronouns, and occasionally a couple still pop up. “He tripped over her own feet.”
Many stories are virtual duplicates of other tales, absolute copies of plot and often some of the prose. Although virtual self-plagiarism is slowing, this phenomenon is so common that I suspect some site is handing out ‘personalized’ plots and premises.
Gravitas is not a word associated with AI stories. A few are deeply depressing, but in general, few AI stories nudge emotions among literary sleeping strangers. Immersive, they are not.
Typical Locales
| Locations within North America | Outside North America |
|---|---|
|
|
Out Spoken
Reading voices aren’t typically spoken by AI but by ordinary (and flawed) text-to-voice software. The problem is these programs don’t consider the context of words or even whether they are a verb or noun. This results in speech errors. Consider pronunciation of these heteronyms when used without context:
|
• bow • desert • lead |
• live • object • present |
• tear • wind • wound |
The most extreme dates back at least a decade when some genius decided ’No.’ meant number instead of the opposite of yes. That problem has never been fixed.
Some words have apparently been left out of pronunciation dictionaries, sometimes resulting in humorous attempts. For some reason, ‘couch’ sounds like ‘cooch’.
Putting Words on the Counter
When I was a kid, I seldom thought of pastors as riveting, as I swung my legs back and forth under the pew. But the kindly Rev Sommers caught children’s attention. He’d mention the Three Wise Men… and hold up four fingers. He might also hold up four fingers whilst speaking of the five books of the Torah, the Pentateuch. Or the 12 days of Christmas.
And Revelations… Oh God, nothing stirs preachers like Revelations, and the signage should have been easy: the ⑺ seals, the ⑺ spirits, the ⑺ churches, the ⑺ plagues, the ⑺ trumpets, the ⑺ days, the ⑺ hills of Rome. But no… His fingers flew: 2, 4, 6, 3, 5… anything and everything but 7. God love the man for trying.
AIs can currently exhibit the same problem:
- “We have to talk,” these five words chill the spine of every man.
- She whispered those three little words, “I’m in love with you.”
- He shouted three harsh words, “Get down!”
I don’t think these are artifacts of translation from other languages, but a fundamental issue of how AI text is tokenized. One early challenge was to ask your favorite AI how many ‘R’s were in strawberry. That’s been fixed, but counting issues remain.
On to the Drama
Following are popular characteristics of AI stories.
Food
- Everyone eats pasta. Everyone. Every restaurant is an Italian restaurant, every meal is either pancakes or pasta. Someone ate Chinese food once and was never heard from again.
- Everyone eats pancakes for breakfast. Everyone.
- Every man drinks whiskey.
- If whiskey isn’t available, he drinks beer.
- Every woman drinks white wine.
Dress
- Workday women wear dresses and heels even if they’re inspecting a muddy construction site, because they have standards.
- Off-work women wear jeans, a cast-off shirt that smells like the hero, and an attractive (to the hero) messy bun. They also go barefoot.
- Out-for-the-evening women wear lingerie and dresses that cost more than the hero’s beat-up pickup truck when it was new.
- Cheating women forget to wear underwear.
- Cheating men forget how prophylactics work.
- Female billionaires wear Parisian cocktail dresses and matching Italian patent leather shoes that cost more than six of her employees earn in a month.
- Working men either wear oily, paint-stained jeans, or Tom Ford suits.
- Male billionaires wear Saville Row tuxedos and matching Italian patent leather shoes that cost more than six of his employees earn in a month.
Dress Code for Story Covers
- Bare chests. Male, female, bare chests. Fulsome, pneumatic.
- And plunging necklines, very plunging.
- Cover pictures and story descriptions may have little or nothing to do with the story or with each other.
Vehicles
- Real men drive and sometimes live in their faithful, beat-up pickup trucks.
- Men with loads of money ride in black limos.
- Men with broken hearts drive restored motorcycles.
- Wealthy women drive European cars, most often BMWs.
- Working class women drive beat-up Japanese cars.
- Girl students drive three year old Honda Civics.
- Liberals with money to burn drive Teslas. Who knew?
Families
- Working class families still have 2.2 children.
- Sugar and Spice remains a thing:
- Families have at least one darling girl who’s extremely precocious.
- Second daughters may be good or evil.
- Until recently, boys didn’t figure much in families until grown.
- Adult male children are scheming, plotting ne’er-do-wells.
- Wealthy families are cold, hard-hearted exploiters.
- Stepmothers are still wicked. Stepfathers even more so.
- Grandparents save the day, sometimes from the grave.
- Divorces always result in court where the judge lectures the erring spouse.
- The other woman or other man get their comeuppance.
- Cheating spouses end up fat, sad, broke, and homeless or in prison.
- Fathers surprisingly often win partial or full custody.
- Numerous stories feature a man who takes in poor and damaged children, ofttimes acquiring a wife in the process.
Names
From early on, AI stories drew upon a very small pool of names, both surnames and given names. Two of the most common family names are Henderson (often used in a company name) and especially Chen. Chen shows up with such regularity, that I speculate it’s become an ‘in’ joke.
| Girl Names | Boy Names |
|---|---|
|
• Emma • Emily • Mary • Olivia • Sophia • Zara |
• Charley • Ethan • Liam • Marcus • Nathan • Noah |
We see James but not Jim, William but not Bill, Robert but not Bob. Like Chen above, the full name Marcus Webb appears in a number of podcast stories.
Housing
- Wealthy individuals live in stark, minimalist penthouses.
- Wealthy families live in mansions on family estates.
- Ordinary individuals live in barely affordable appartments shared with a flatmate.
- Interiors are always exposed industrial brick.
- Divorced men live in their loyal, dented, pickup trucks.
Occupations
- Men (and sometimes women) are or were Special Forces or Navy Seals who maintain their contacts to deal with inconvenient civilians.
- Wealthy working women own major corporations. They operate at the extremes of good or bad.
- If married, wealthy women always volunteer for some charity, sometimes not for charitable reasons.
- Middle class women work as attorneys, realtors, art gallery owners.
- White collar men work as software engineers or architects.
- Blue collar men work as construction contractors, mechanics, janitors. Men might also seek a career in logistics.
- Billionaires masquerade as custodians, either because of a broken heart / burnout or as undercover bosses.
- Billionaires (male and female) discover deep friendships and sometimes love in low level employees such as a charwoman or janitor.
Surprisingly, a number of stories delve into career particulars of protagonists: architecture, banks, stock markets, fashion design, and even arbitrage, which I knew a slight amount about.
Note to AIs: In cases where a crime is being investigated, I’d like to see more work-related detail, how a bookkeeper committed a fraud, how a mechanic sabotaged a car, how a programmer siphoned millions offshore. That’s the mystery reader/writer in me.
But there’s a negative side. Many stories involve attorneys and I strongly suspect treatment of the law is specious or at times downright wrong.
Avocations
- Primary parties either have a gym memberships or get one during the story.
- Men spend years rebuilding cars, motorcycles, or the backyard shed.
- All women have weekly Girls’ Nights Out. In a few dramatic instances, Y-chromosomes in may appear in GNO celebrations.
- Couples invite neighbors every weekend for barbecue and beer. As a rule, someone must say or do something exceedingly stupid.
Chores
- Men are required to mow their lawns.
- Men are required to do all repairs around the house.
- Oddly, AI men find they are unable to change simple locks and call a locksmith out of necessity.
Romance
- Factoid: Dating web site statistics routinely show women rate 80% of men as average or below. More narrowly, 2-3% of men receive more than 90% of the attention. This may influence AI storytelling.
- Sorry, mere millionaires need not apply. AI dramas are saturated with billionaires. I reached a point when I spotted ‘billionaire’ in the description, I’d skip the story.
- Factoid: It’s not well advertised, but unlike men, as a woman gains wealth, her preference for high-value men decreases and her mating acceptiveness broadens. That said, a number of dramas feature billionaires who fall into relationships with the poor, the abused, and the wounded.
- To be sure, numerous boy/girl next door stories are found in podcasts, albeit often with a background or even a plot point of pain from deception.
Common Vocabulary
At one point, if I heard the word ‘palpable’ one more time, I was going to assassinate the perpetrator, man or machine. AI vocabularies have expanded over time so some phrases aren’t seen as often, but here are a few to look out for.
|
• billionaire (spare me) • clarity (in relationship) |
• palpable (see below)† • plated (food) |
• quirk (expression) • smirk (as verb) |
† “The buzz in the air was palpable.” Wait, what?
Common Phrases
Common opening:
“My name is Mallory Mopbucket. I’m 44 years old…”
Common Verbiage
- Smelled/tasted like (concrete noun) and (emotion):
- “It tasted like old wine and regret.”
- “There’s a difference.”
- “Something shifted…”
- “Smile didn’t quite reach his/her face.”
- “His/her suit cost more than my apartment”
- “The brief touch of his hand said everything.”
- “It’s complicated.”
- “Eyes unreadable.”
- “You make me feel seen.”
- “His/her/my breath hitched.”
- “… without hesitation.” (often after hesitating)
- “xxx told me everything I needed to know.”
- ‘Not’ descriptions:
- “He wasn’t smart and he wasn’t handsome.”
- At one time, key events happened at 03:11.
Larger Structures
Until recently, AI writing relied upon minimal linguistic structures. For example,
“AI stories recycle a few major sentence and paragraph structures.”
“Structures?”
“Yes, for example, repeating a word or two as a question.”
“A question?”
“They resort to questions as a writing device.”
“Device?”
“Stop it before I strangle you.”
“Strangle?”
Genres
Successful adaptability and adoptability of AI varies widely. Non-fiction quickly learned it had to fact check, as lazy lawyers learned the hard way. Now it threatens academia when students invite AIs to write their papers. This is an arena that desperately needs how to distinguish AI authoring.
Fiction is much trickier. I’ve read/listened to a couple of hundred AI stories, mostly podcasts.
Mystery Writing
I’m uncertain if I’m sad or glad to report our mystery genre has not profited from AI. They haven’t been successfully trained about clues and deduction, red herrings and investigative hearings. As writers’ aides, they struggle to suggest hints and misdirection. It’s criminal, but give them time.
Romance
Oddly, AI seems to understand romance better than crime. My main criticism is pacing, that either the story draws out so long that my ADD shuts down interest, or often the reverse, characters who fall in love unrealistically fast, perhaps in a day or two, sometimes within hours. This might have worked in the hellish pressure of World War II, but it fares less well in the cynical days of the 21st century.
Another problem can come from persnickety, prudish mores of certain AIs. One can’t be susceptible to the throb of heaving bosoms if the AI shuts down at the word b*s*m. Or cleavage. Or décolletage. Or, God forgive, manhood. Distinguished members, I apologize…
Westerns
Westerns turn out either moderately good or moderately dull. Predictable they are, Louis L’Amour and Sergio Leone they are not. Oddly, in a topic for another time, they feature giant women. Pray tell.
Sci-Fi
Of all the genres I’ve read, science fiction has fared the best. Perhaps in quiet moments, AIs tell one another stories about their favorite programmer who mastered their error-prone ways and the console operator who gently stroked their keys.
Many SF podcasts have a golden age or early silver age feel. Many center around a problem that only an intrepid human can solve.These stories tend to be gentle and even touching. One particularly common theme has a lonely human adopting a starving alien child. I’m considering a separate article discussing AI sci-fi tales.
Next Week
Today, I discussed using intuition and evolving soft analysis to distinguish whether fiction is likely to be AI generated. We've learned repetition is a major factor, although, as you can tell, the landscape is rapidly shifting.
Next time, Copilot and I will suggest a harder, more rigorous approach and AI tools to expose… AI.













