8.4.2024

A special edition of EP . . .

As usual, I had planned a different post for today (it’s the one that keeps getting nudged out of line!)—but then I saw this post from the inimitable Frederick Woodruff:

I read it immediately, of course. And was intrigued in a general way.

Then I got to this part:

Wow, indeed . . .

I don’t usually bring current events or politics into EP, mainly because I don’t want the content landscape to be any more sprawling than it already is! But I couldn’t resist finding out whether the Tarot has something more or different to say on this topic. And since it’s widely assumed that Harris is in the final stage of deciding on her VP choice, today seems the proper time to ask.

So I took this opportunity to further test my curiosity about whether/how Large Language Models might function as a crowd-sourced divination tool.

If you didn’t read earlier posts where I talked about this—and recounted my own unexpected experience—now’s the time!

Here’s the flash version:

I tried out a prompt chain designed by someone else to produce a Tarot reading. I used two different premium models (ChatGPT4 and Gemini Advanced), and found the results interesting in a general way. I considered the whole thing an abstract experiment.

But the next day, one of the “readings” seemed to connect with events that took place in my life. Since then, I’ve intended to repeat the experiment—but didn’t want to be the subject this time. So when I read Woodruff’s post, the opportunity seemed irresistible.

And I have to tell you . . . the result is fascinating.

Here’s what I did:

• Used only Gemini Advanced, which I thought gave better results in my original experiment.

• Used the same prompt chain—but in this case, I added a single word at the beginning of the prompt to serve as a sort of virtual querent.

• Ran the experiment seven times: once without a keyword, then once for each of the six known contenders, using just their last names.

• Analyzed the results, and compiled the whole thing in a document. The contenders are divided into two “tiers,” based on public indications.

I’ve created a PDF of the document—and you can download a copy if you’re interested. It contains the unedited responses provided by Gemini Advanced.

Vp Readings

198KB ∙ PDF file

Download

Download

And now . . .

Here’s an overview report, with some notes:

Noteworthy:

1.      Two contenders (Kelly and Shapiro) share the World card with Harris.

2.      Those three (Harris, Kelly, Shapiro) also have a Pentacles card.

3.      In fact—all but one (Walz) of the six contenders has a Pentacles card.

4.      Only two contenders (Kelly and Shapiro) have Major Arcana cards.

5.      Only one (Shapiro) has two Major Arcana cards—and they are 1 and 22.

Statistically Speaking

• Looking at the whole 21-card draw, the number of trumps is a little low—just 4 for 21.  An average number would be 5 or 6.

• The proportion of Pentacles—6 for 21—is significantly above the average, which would be 4.  Wands come in at 5, and Cups at 4, while Swords lag far behind with just one occurrence.  Aces are double the average, which would be 1 for 21.

• Pritzker has a statistically unusual result of 3 pips from different suits, with the same value.

Observation

I intended the “no keyword” draw to reflect the overall situation . . . .

But it’s tempting to see the Queen of Wands as Harris herself—especially since there are 3 Queens in the group of 21.

What are the odds? Out of every one hundred times you do a 21-card tarot spread, you can expect to get three instances of the same court card only once or twice.

---

So that’s how my experiment turned out! I wanted to share the results before whatever happens next happens.

I’ll wrap up by noting that the two contenders who stood out so strongly in Woodruff’s inquiry—Shapiro and Kelly—were the same two who shared the World card with Harris.

I’m guessing that reflects the dynamics of the situation on a deeper level than we’re likely to know. Not necessarily the outcome, but perhaps some of the archetypal structures at work.

More very soon, C