Pumpkin Spice and the AI Apocalypse

Ah, autumn in America. It’s time for some of our favorite seasonal rituals: leaf-drenched pumpkin patches, foamy lattes, retro barn jackets, and yet another self-reported Anthropic AI researcher declaring to the world that the product he’s been working on might just murder us all.

The latest AI alarmist

It’s almost become charming, these occasional alarming cries sent over the Internet, shouting that the technological end is inevitably near. They harken to seasonal marvels like the annual ruby-throated hummingbird migration, except imagine the hummingbirds depressed, demoralized, flying through robot shrapnel, and constantly texting their therapist.

Even better, like many things buzzing around the Internet these days, it’s hard to tell whether these apocalyptic warnings are even real. On Wednesday, the latest AI warning went out over an X account from Jacob Coxon:

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately. No other human activity poses this level of danger.

At press time, the posts had picked up millions of views.

Coxon’s X handle is @hilbertspaess, which immediately made me suspicious, because at first I found it eerily close to one of those weird, seemingly made-up brands that pop up when you try to buy cheap athletic workout gear on Amazon for your high school son. Here are some examples of brands selling shorts on Amazon that are 100% real: “HQUEC.” “BALENNZ.” “COOFANDY.” “CANGHPGIN.” And finally, my personal favorite: “RDRUKO.” It’s my favorite for the obvious reason, which is that it spells “OKURDR,” but backwards.     

However, because I am a true professional, upon further research – research NOT USING AI, thank you, although I definitely probably did accidentally use AI without even knowing it, because I was using a computer – I found that “Hilbert’s pass” is actually an AI framework term, which you can Google if you’d prefer to be bored to death instead of eventually hunted down by future Terminators.

Another one!

Anyway, another AI guy who is apparently the “Alignment Science lead” at Anthropic – again, as a severely tech-challenged individual, I seriously have trouble believing any of this is happening in real life – weighed in on Coxon’s post on X, agreeing that “we really do earnestly believe AI could kill all humans! I personally think it is >10% in the next decade.”

Ah. Well, that’s only a 1 in 10 chance of societal destruction. Proceed apace, friends. Full speed ahead.

But seriously: If this is all true – and it could be, but it could also be a hoax, or part of a viral marketing campaign, or written by AI itself – what in the heck is a normal person supposed to do? You could get an off-grid cabin in the middle of the Tennessee woods, but that’s a lot of work. You could give up, quit your job, yell “YOLO!” and head off to Sardinia. But that costs a lot of money, and then when AI hacks your bank account, you’re not only stuck in Sardinia, you’re poor. You could conscientiously object to this whole bit of nonsense, but you really can’t, because AI is everywhere, and it’s probably watching you from your toaster right now.

Countervailing perspective

I’m not completely innocent. When it first came out en masse, I did try to use AI once a few years ago, attempting to recreate a digital image of a dream I once had involving “Motown Song,” which is a legendary piece of tonal genius, with Rod Stewart featuring the Temptations. The dream featured a full musical montage at a random McDonald’s, ending with Rod Stewart jauntily tossing me a mandarin orange over the checkout counter. Back then, AI refused my request. I suspect trademark law was to blame.

I also admit to owning a robot – two robots, in fact, off-brand Roombas that are probably judging me as I speak. I have these fake Roombas in my house because I also have a giant 75-pound German shepherd whose shedding could personally provide the supplies to construct two full mini-goldendoodle replicas per hour.

Do you know what’s great about dogs, by the way? They don’t have QR codes. They don’t ask you for your password sixteen thousand times, or try to charge you $100 to talk to them on the phone, or insist you get the equivalent of a full eyeball scan so you can log into an account you never wanted to make in the first place that then won’t allow you to cancel a simple car reservation at a company that theoretically may or may not start with an H and rhyme with Schmertz.

Ahem.

An ironic twist ending to this AI story

Last week, the beloved writer Wendell Berry, who treasured the agrarian life and the natural world and tended to eschew technology, passed away. My husband texted me a few lovely articles memorializing his life and work – and then, as an addendum, he added an in-depth PDF that compared and contrasted Berry’s work to that of Paul Kingsnorth, a writer whose recent book, “Against the Machine,” warns against the ill effects of our tech-obsessed world.

You guessed it – that PDF was made by AI. I deleted it in a righteous huff, so it’s now gone.

The irony, however, lives on.  

This article was originally published by RealClearPolitics and made available via RealClearWire.