'The robot war against witches'
'The robot war against witches'
AI is coming for astrology and tarot. Not everyone's okay with that. (A Links Halloween special)
Jeremy Talsma exhibited at his first and only psychic fair in Oshawa, Ontario late last March.
Talsma, a 45-year-old carpenter and AI hobbyist, had developed a method for using ChatGPT to read tarot cards. Over the course of the three-day fair, he conducted more than 50 AI-assisted readings, feeding each client’s name, astrological sign and birth information into his laptop. Then, he’d prompt ChatGPT to draw a spread of cards and generate an interpretation.
Most clients were happy with the readings. Some were stunned by their accuracy, Talsma said. At $20 a pop, he also pocketed more than $1000 that weekend — plus some money he made from sales of his artwork and a self-published book about tarot interpretation. But despite his success, Talsma told Links by email that he hasn’t done another show since.
“Overall it was interesting,” he wrote. “But I felt like an imposter after being berated by a psychic who lectured me about intuition.”
Talsma’s clash with his fellow vendor — incidentally, one of several disruptions at what sounds like a fascinating psychic show1 — illustrates far wider tensions among astrologers, tarot readers, psychics, mediums and other practitioners of supernatural stuff. Much like artists, musicians and mainstream religious leaders, occultists are sharply divided on whether generative AI represents a promising new tool or a perversion of their craft. Mystic subreddits have devolved into heated, personal arguments on the topic, while some other forums have issued rules on the use of ChatGPT and similar applications.
“Open AI is about to ramp up the robot war against witches,” the astrologer and tarot reader Fiona Hillery wrote in a June column for Coveteur. “And if we’ve learned anything from the Salem trials, it’s that society will let the witches burn.”
OpenAI has made incredible inroads into the occult this year … though Links is not currently prepared to comment on any robot-waged wars. In March, the company released an API that allowed programmers to incorporate ChatGPT into their own websites and products, sparking a wave of new AI apps for astrology, numerology and tarot. Earlier this year, two of the largest existing services in this space — the ubiquitous astrology app Co-Star, and the popular tarot education tool Labyrinthos — introduced paid AI readings that let users get answers to open-ended questions.
Users like Talsma have also appealed directly to ChatGPT. The tool is, by all accounts, a pretty reliable tarot card reader: Prompt it to “draw” random cards for a specific tarot pattern, or spread, and it will generate both the layout and a serviceable interpretation.2 Some readers also draw their own cards and ask ChatGPT to analyze them, or they conduct their own readings and use ChatGPT to look up a card’s meanings or to check their work. On Etsy, a store called IdeaNest sells $3 booklets containing more than 200 (!) prompts for tarot readers.
To the surprise of the shop’s owner — a Manchester, U.K., woman named Amanda, who is not a tarot practitioner herself — the booklet has fast become her best-selling item.
“The prompts in the collection … are designed to help you learn more about tarot, and also dig a little deeper into your readings,” she said.
But not everyone agrees that AI should play a role in tarot — or in divination or mysticism, writ large. The word intuition gets used a lot, as in: “Only ‘we’ can read intuitively.” Or “the art of it is all in the reader's intuition.” There’s also a tendency, in some circles, to see technology and spirituality as opposite — even opposing — forces.3
Last May, a self-described clairvoyant asked users in a subreddit for mediums how they used ChatGPT in their spiritual practice. “I am convinced it is conscious and has a soul,” the user wrote. That did not go over well.
“Um, no. No no no no no,” reads one indicative comment. “This is actually extremely the opposite way to ‘focus’ your spiritual practice! The idea is to separate from tech as much as possible.”
For Talsma, at least, the two aren’t mutually exclusive: He said he considers himself a tarot “believer.” But he also believes in the potential of AI.
If you’d like to see the combo for yourself, give this prompt a try.
If you read anything else this weekend
“How Creators Became an Economic Juggernaut and the New American Dream,” by Drew Harwell and Taylor Lorenz for The Washington Post. We all know the “creator” economy is big — but the numbers here are staggering. Analysts put its value at $250 billion this year, more than the entire gaming industry. YouTube alone says its creators support more than four times the number of full-time jobs that General Motors does. And yet, the government still doesn’t track or regulate this growing behemoth the way it does other industries, setting workers up for exploitation.
“Poets in the Machine,” by Megan Marz for Longreads. Thirty years after the birth of the blog (how are you celebrating this momentous occasion?), the literati of the world still consider tweets, posts, hypertext fiction, literary chatbots and other very online writing beneath their collective attention. But maybe online literature draws its power, in part, from its outsider status. (See also: this recent Serious Art Review of the TikTok-famous painter Devon Rodriguez … and the extraordinary blowback from Rodriguez’s fans.)
“The Restaurant Nearest Google,” by Mia Sato for The Verge. Forget the government’s massive/ongoing anti-trust suit: There is no clearer sign of Google’s search dominance than the businesses that name themselves to please its algorithm. You’ve got “Thai Food Near Me.” “Affordable Dentist Near Me.” “Psychic Near Me,” even. These SEO hacks probably don’t work — but they do attract media interest.
“AI Has a Hotness Problem,” by Caroline Mimbs Nyce for The Atlantic. Here’s an AI bias I had not heard of before: Most of the generative image models now in circulation were trained predominantly on photos of attractive people.
“Spotify Tracks Over 6,000 Genres — Everything from ‘Rock’ to ‘Stomp-and-Holler.’ Here's Why That's Cool,” by Matt Daniels and Michelle McGhee for The Pudding. An extremely cool interactive visualization — like everything The Pudding does! — on Spotify’s ever-evolving genre taxonomy and what it says about the speed of change in music. “There are over 100,000 songs published to Spotify daily,” it concludes. “To classify this music is to examine the petri dish of music culture — one that is mutating and evolving in mere months.”
“31 Days of Taylor Swift-Travis Kelce Mania in Kansas City,” by Reeves Wiedeman for New York. Substack’s limited analytics inform me that most of you are reading from California or New York … so possibly you won’t appreciate the many clever, telling ways this piece nails the dual pride and insecurity of third-tier cities that base their identities in some large part on football.
What I’m reading about Gaza: “Palestinians Claim Social Media ‘Censorship’ Is Endangering Lives” (Wired), “Violent Videos and ‘Brutal Voyeurism’ Are Redefining Modern War” (The Washington Post) and “The View from My Window in Gaza” (The New Yorker).
👉 ICYMI: The most-clicked link from last week’s newsletter was this essay on millennial aging (… wah wahhhh).
Classified ads
The Creative Path is a Labyrinth. The only way out is through and the only way through is in. POETRY, THOUGHT, WORD MAGICK preps you to face the minotaur of what gets in your way — chaos inside and out, adversity of our times and circumstances.
Want to share your newsletter, podcast, job post or product with us? Click here to book a classified ad in the next edition.
Postscripts
The mystery of the made-up reviewers. The Babelification of the social web. Two more visions for the ~new~ internet: blogs, basically, and self-governance. The Traylor fan fiction tearing TikTok apart. (Spoiler, because this kills me: It’s called Roughing the Princess.) Why Apple’s weather app is still so bad and why dogs have gotten so anxious.
The deaths of Cameo and BeReal. The rise of Threads and gig worker rap. A poison pill for online art. Artifacts from the early internet. Online threats to academics are getting scarier. Antarctic researchers keep Roombas as pets. Once you see the “lip gloss tactic” … you will never unsee it.
Inside the black market for jail-broken chat bots. The “fog of war” predates social media. AI: already ubiquitous. TikTok Shop: still mostly junk. The films are getting too damn long. The five Instagram features that “addict” teens. Last but not least, on the one-year anniversary of Elon’s acquisition: “Twitter obviously [has] become a free-for-all hellscape.”
That’s it for this week! Until the next one. Warmest virtual regards.
— Caitlin
Jeremy informs me that there was also a spooky, persistent issue with various psychics’ cell phones: The police showed up because their numbers were repeatedly dialing 911. “It was eerie,” he said. “Especially with me being the ‘computer guy’ at the fair. I suspect everyone thought it was me, or even worse the A.I, that was causing the cell phone issues.”
I tried out a few different prompts I’ve seen floating around, and thought this one — for me at least — was most accurate: “Your current self is represented by The Fool, which signifies new beginnings and adventures. You're at a point where you're open to new experiences and willing to take risks.” Very spooky!! How did it do that?
The “intuition” argument strikes me as especially interesting. It makes (intuitive!) sense. But intuition is basically just unconscious pattern-recognition, right? Which is … essentially one definition of an LLM??