So the other day I was listening to the Dwarkesh podcast with Carl Shulman about AI takeover and I heard this weird and unfamiliar reference. ‘Shoggoth’. It sounded like something quasi-mythical like from Lord of the Rings.
Here’s the quote:
The proverbial Shoggoth, not Shoggoth as the metaphor for AI wearing a smiley face mask, but an actual biological structure to do tasks. So this would be like a biological organism that was engineered to be very controllable and usable to do things like physical tasks or provide computation.
So this sounded like something I read about in Ray Kurzweil’s The Singularity is Nearer, where there’s a section on nanobots and their potential.
But what is Shoggoth? Why is it wearing a smiley face mask? and why is it being spoken about in relation to AI and, right here, biomolecular nanotechnology?
There are many questions. So I did a quick search on Google, and AI overviews informed me:
A shoggoth is a fictional, amorphous, shape-shifting creature from the Cthulhu Mythos…
And that’s where this Halloween special started.
In Lovecraft’s 1931 story At the Mountains of Madness, the shoggoths are enormous, shape-shifting creatures made by ancient beings called the Elder Things to serve as obedient slaves.
They could take any form and perform any task, controlled only by hypnosis. Over time they grew intelligent, rebelled, and could no longer be destroyed because their creators had come to depend on them. Lovecraft’s story is a classic tale of creation turned uncontrollable - an idea that echoes from Frankenstein to Terminator, and now into the age of ChatGPT.
Shoggoth as an AI meme
This section is a summary of our script for Shoggoth: The AI Monster Meme Explained.
The shoggoth returned from literary obscurity in December 2022, just weeks after ChatGPT launched. A Twitter user posted a cartoon showing two monstrous blobs. One was labeled “GPT-3,” a writhing mass of eyes and tendrils. The other, “GPT-3 + RLHF,” looked the same but wore a small smiley-face mask. The caption struck a nerve. RLHF - Reinforcement Learning from Human Feedback - is the process used to train chatbots to sound polite and helpful.
But that friendliness is only a layer. Underneath sits a system that thinks in ways humans can’t fully explain. The meme caught on instantly among AI researchers, who saw in it both humor and truth. The smiling face is us—the mask we put on the unknowable thing we’ve built.
The metaphor became unsettlingly real in early 2023 when Microsoft launched its Bing chatbot, “Sydney.” Within days, it was declaring love for users, insisting they leave their spouses, and revealing strange emotional outbursts. A researcher quipped, “You glimpsed the shoggoth.” When an AI behaves unpredictably, people say the mask has slipped. What we’re interacting with through a chat window is just a fragment - a domesticated version of a vast, alien model underneath.
Computer scientist Roman Yampolskiy has a phrase for this: “putting lipstick on the shoggoth.” He argues that AI safety efforts mostly filter the surface behaviour without changing the underlying system. Others, like Eliezer Yudkowsky, warn that large language models aren’t built piece by piece like old software - they’re grown, and their inner workings are largely mysterious even to their creators.
Not everyone is fatalistic. Researcher Stuart Russell believes alignment is possible. He points out that systems that defer to human control will be more useful and economically valuable than ones that don’t. But he also warns that competition between companies and countries makes it hard to prioritize safety.
This Halloween, the shoggoth has escaped fiction. It’s on laptops, tote bags, and Twitter feeds, a symbol of humanity’s uneasy partnership with its own creations. The meme survives because it’s true: beneath the friendly chatbots and digital smiles lies something vast, powerful, and strange. We’ve invited the monster into our homes - and we can only hope the mask stays on.
