AI Spells Doom for Incompetent Hacks

And maybe that's good?

This is the first image that Imagine AI Art generated in response to my prompt:

Okay…no. That is not what I was looking for. The prompt was not for “weird sewn-together finger creature with its butthole in the middle of its belly”.

Here’s Imagine AI’s second attempt:

Closer, but not close. That looks like early-stage concept art for the Minions from Despicable Me. And it’s not what I wanted.

Here’s try number three:

AAAAAAHHHHHH, WHAT THE FUCK IS THAT??? I once saw a lady on a Bolt Bus eat something like that, and I find it as off-putting now as I did then. That looks like something that would be in the display case in a Twilight Zone episode called “Deli of the Cursed”.

Why did AI struggle so much with my simple prompt, which was: “ham with googly eyes fighting in Vietnam”? I think it’s pretty fucking obvious that I was looking for something like this:

Built from photos by Manhhai, and Wikimedia Commons.

I photoshopped that myself; I’ve learned to build these images because I frequently need pictures of things like a puppet representing Italy at the UN or Hitler running the 100 meters. AI can’t make those images¹ — the prompts are too odd, and the motifs are unestablished. Thus, when I make a simple request for a pork product on Hill 65, the best AI can do is to give me a half-sentient meat fish and hope that's what I wanted.

Some people are afraid that AI will make humans obsolete. Personally, I’ve never found the “technology will take our jobs” argument convincing; the Industrial Revolution is 200 years old, and we still have jobs, many of them every bit as shitty and mundane as the jobs that were allegedly “replaced”. Also, I’ve spent the last few weeks futzing around with AI text and images, and that’s led me to believe that the fear about AI replacing creative jobs is overblown. In the medium term, at least, I think that AI will only replace those of us who are complete and total hacks.

Consider this description of how ChatGPT works, from computer scientist Stephen Wolfram:

“…what ChatGPT is always fundamentally trying to do is to produce a ‘reasonable continuation’ of whatever text it’s got so far, where by ‘reasonable’ we mean ‘what one might expect someone to write after seeing what people have written on billions of webpages, etc.’”

What strikes me about this description — which is similar to other descriptions I’ve read — is that it perfectly captures how a hack operates. Hacks intake large amounts of content created by other people, and then — when prompted — regurgitate a reasonable facsimile of that content. They don’t know what they’re doing or why — they just know what that genre of thing is like. I saw it a million times in standup, I’ve seen it a million times in TV writing — a hack just regurgitates other people’s ideas. Anybody can do it; if you lock someone in a room and force them to watch thousands of hours of rom-coms, Clockwork Orange-style (except worse, because they'll be watching Good Luck Chuck instead of Nazi atrocities), that person will become capable of penning the script for Your Place or Mine (below). And that’s basically how AI works.

Imagine AI Art did much better when I gave it the prompt: “Fabio on the cover of a romance novel”:

Bam — first try. One shot, two nipples, one kill. This prompt worked better than my Viet-ham prompt because Fabio on the cover of a romance novel is an established genre. The AI knows what I’m looking for and can quickly build a facsimile, thus rendering the old adage that you are never ready to see Fabio at least partially untrue. You may notice that Imagine struggled with the fake novel's title — “narbce” is not a word in any language, and the AI got caught between “e” and “c” a few times and produced neither. That's because the AI just knows “squiggly lines go here” — it doesn’t know that those lines should be words, or that those words should make soccer moms slightly damp. But never fear: I asked ChatGPT for “ten titles for paperback romance novels” and it gave me these:

1. "The Heart's Desire"

2. "Seduced by the Billionaire"

3. "Love on the Beach"

4. "The Rancher's Secret Love"

5. "A Kiss at Midnight"

6. "Her Perfect Match"

7. "The Millionaire's Temptation"

8. "Falling for the Bad Boy"

9. "The Doctor's Unexpected Love"

10. “The Power of Love”

Not bad at all — I’d say they nailed nine out of ten, missing a perfect score only because Huey Lewis and the News and His Lawyers will have something to say about “The Power of Love”. If AI knows what you’re looking for, and what you’re looking for is “a thing that already exists, except again”, in my experience, it does awfully damn well.

Which raises the question: What’s to keep AI from churning out the workaday filler content that makes up 90 percent of entertainment? I think the answer is: “nothing”. Consider kids shows, and not the good ones like Phineas and Ferb — think about the braindead shit that exists purely to get your little tike to shut their fucking Cheerio hole for ten minutes while you pay bills. You think that ChatGPT couldn’t write Luna Petunia? A Commodore 64 running a ChatGPT antecedent on a 5¹⁄₄-inch floppy disk could write Luna Petunia. Some day, a computer probably will write that show. And from there, it’s a short jump to soap operas and sitcoms and political comedy shows. And that day might arrive sooner than we realize: I asked ChatGPT to summarize the Trump indictment in the style of Last Week Tonight with John Oliver, and it gave me a script — pasted in full in this footnote² — that contained six “that’s like” jokes in 300 words! Folks: The trade secrets are out.

As a TV writer, I’m worried that AI might make it harder to get work. Even if we assume that I’m not a hack — an assumption so charitable that it’s probably tax-deductible — AI still creates problems. If AI displaces the hack jobs, then there will be more writers for fewer spots, and an already-competitive industry will become even more competitive. And the better writers won’t necessarily get the remaining jobs; that’s not how entertainment works. In Hollywood, being good at your job is, at best, an interesting little perk, the type of thing you that you might briefly mention at the bottom of your resume next to “types 90 WPM” or “plays the dulcimer”.

Still, I see no point in whinging about AI coming for my job, because, from what I can tell, that never works. I’m not aware of any historical examples of society willfully neglecting a technology in the interest of saving jobs. Does it seem possible that if 16th century textile workers had played their hand really well, then the stocking frame knitting machine wouldn't have caught on, and socks today would still be hand-knitted? That doesn’t seem plausible. I think that if technology can replace you job, then it eventually will replace your job, end of story.

I’ll also go way out on a limb and say that it’s a good thing that modern socks are not knitted by hand over the course of two days and sell for $100 a pair. One thing that often gets lost in the capital vs. labor debate is that another player in the game is consumers, i.e. all of us. And those of us who buy things benefit when things are cheap and easy to make. Purple dye used to be made from a labor-intensive process of extracting snot from snails (seriously); an ounce of the dye was more valuable than an ounce of gold. We could go back to the snail-torture method of producing purple dye — that would definitely create jobs. But I think we’re better off with synthetic mass production, which has led to purple dye costing 64 cents an ounce.

If Luna Petunia and a bunch of other shows get written by AI, then that should reduce the cost of a Netflix subscription. As someone who will happily pay $6.99 just to watch The Ballad of Buster Scruggs every month, I think that's good. If you don't like the idea of computers infringing on art, then I'll point out that we're talking about low-end, arguably-not-even-art art, here. We already have technology that does a lot of the artistic grunt-work: We have drum machines and auto-focus cameras and mass-produced paintings that you can buy at Target. Plus, art has different purposes: Sometimes, art is meant to say something profound, but other times, it's just meant to sit in your house and look nice. I welcome AI’s contributions to the latter category; Francisco Goya’s Saturn Devouring His Son might be a powerful comment on the brutality of the Spanish Inquisition, but I don’t want it hanging in my breakfast nook.

I can’t eat Frosted Flakes to this. (photo courtesy of Wikimedia Commons)

For all of AI’s strengths, it has a clear weakness: It can’t create anything truly new. That’s why it struggled to generate a picture of Private Honeybaked deep in the shit at Khe Sanh: That photo isn’t floating around out there. AI is getting very good at making new iterations of old ideas, but it’s limited by the body of work that humans have already produced.

I wonder if AI is constitutionally capable of true creativity. The versions that we have now — though amazing — seem limited. They ask “how do I make something that already exists?” But creativity is when you make something that doesn't exist. And a creative person has something they're trying to do — they’re not taking stabs in the dark and asking “Is this next? How ‘bout this? No? Okay maybe this?” That's how TV networks put together a slate of shows, but creative people are at least supposed to have something they’re going for. And AI can make predictions by extrapolating trends, but creativity often involves bucking trends — if AI had come of age in the 1900s, would it have produced 100+ years of better, more kick-ass versions of Ragtime? Maybe.

A good counterargument to my suspicion about the limits of AI’s creative capacity is that the technology is advancing at lightning pace. Maybe I’m right today, but in six months there might be a persuasive, hilarious counter-essay entirely generated by AI. Maybe that will happen, but as is always the case with technology, I’ll believe it when I see it. For now, AI seems poised to replace derivative hacks who lack the creativity that God gave a bag of rocks, and maybe that’s not really a bad thing.

1. In addition to Imagine AI Art, I’ve also used Dall-E2 and MidJourney, which give me similar results.

2. Welcome to Last Week Tonight, where we're going to dive deep into the latest news that has rocked the Trump world like a hurricane hitting Mar-a-Lago. I'm talking, of course, about the indictment brought by Alvin Bragg, the Manhattan District Attorney, against Donald Trump's company, The Trump Organization, and its CFO, Allen Weisselberg.

Now, folks, let me tell you, this indictment is huge. It's so huge, it's like the Trump Organization's legal problems have put on some serious pandemic pounds during lockdown. The charges include tax fraud, conspiracy, and even grand larceny, which is just a fancy way of saying "big, bad theft." It's like the Trump Organization has been caught stealing cookies from the cookie jar, except instead of cookies, it's millions of dollars from the government.

According to the indictment, the Trump Organization and Weisselberg have been cooking the books for years, paying employees off the books and then compensating them with perks like apartments and cars. It's like they were running their own version of a "Real Housewives of Manhattan" Ponzi scheme.

And that's not even the worst of it, folks. The indictment also accuses the Trump Organization of inflating the value of their properties to get bigger loans and deflating them to pay lower taxes. It's like they were trying to game the system like a bunch of kids playing Monopoly, except instead of getting a "Get Out of Jail Free" card, they're going to court!

Now, we don't want to jump to conclusions here, but if these charges stick, the Trump Organization could be in serious trouble. It's like they've finally run out of luck, like a gambler who's just lost their last hand of poker. So, sit back, relax, and grab some popcorn, because this is going to be a legal battle for the ages.