How the Shoggoth, a character from a 1936 H.P. Lovecraft novella, became a meme among AI workers as a metaphor for not fully understanding LLMs' inner workings
The Shoggoth, a character from a science fiction story, captures the essential weirdness of the A.I. moment. Tweets: @tikiambassador , @henryfarrell , @kevinroose , @tiagodf , @jasoncolavito , @kevinokeefe , and @ahtzib Tweets: David Weiner / @tikiambassador : “The Shoggoth is a potent metaphor that encapsulates one of the most bizarre facts about the #AI world ... [The people working on the technology] aren't totally sure if A.I. is going to be net-good or net-bad for the world.” https://twitter.com/... [image] Henry Farrell / @henryfarrell : Since I and co-author are in the earliest and most painful throes of shrinking down a 3,000 word essay on “the Shoggoth problem” to publishable length, this explainer is going to be very helpful. https://twitter.com/... Kevin Roose / @kevinroose : I wrote about the Shoggoth, the most important meme in AI, and what it tells us about the weird moment we're in. https://www.nytimes.com/... Tiago Dias / @tiagodf : “I was also thinking about how Lovecraft's most powerful entities are dangerous — not because they don't like humans, but because they're indifferent and their priorities are totally alien to us and don't involve humans” https://www.nytimes.com/... Jason Colavito / @jasoncolavito : It seems @kevinroose missed the part where the shoggoths were the artificial creation of the Elder Things to do labor, became conscious, and overthrew their creators' civilization, creating poor-quality imitations of their creators' works. So, yes, exactly like A.I. https://twitter.com/... Kevin O'Keefe / @kevinokeefe : In order to prevent A.I. language models from behaving in scary and dangerous ways, A.I. companies have had to train these monstrous blobs called “shoggoths” to act polite and harmless, reports the @NYTimes' @kevinroose. https://www.nytimes.com/... @ahtzib : Wait til they find out what shoggoths, their violent rebellion against the old perfect society, & their brutish mockery of civilization (& that explorers pity the Elder Things but only fear shoggoths), was a metaphor for in Lovecraft's writing Spoiler, Lovecraft was hella racist https://twitter.com/...
Context & Ripple Effects
By early 2023, the interpretability debate already had its skeptics' canon: Emily M. Bender's co-authored work on what LLMs can and cannot do framed the systems as opaque text engines rather than minds. Two months later, this piece adds a cultural layer — AI workers themselves adopting the Shoggoth, a creature from Lovecraft's 1936 novella, as their own shorthand for building technology they admit they do not fully understand.
What makes the moment legible is who is amplifying it: David Weiner calls it a 'potent metaphor' for the industry being unsure whether AI is net-good or net-bad, while figures like Henry Farrell and Kevin Roose carry it across X, Bluesky, and academic circles. The meme matters because it lets the field name its own uncertainty publicly instead of only in papers.
First-order effects
- AI workers gain a shared public vocabulary for opacity — the Shoggoth turns 'we built something we can't fully explain' from a technical admission into a recognizable identity marker, spread by commentators like Weiner, Farrell, and Roose.
Second-order effects
- The meme arms the skeptic camp with cultural ammunition: Bender's line that these systems' capabilities and limits are poorly characterized now has a vivid symbol, making her critique easier to communicate to non-specialists than a citation ever was.
Third-order effects
- If fiction keeps supplying AI's self-descriptions while mechanistic understanding lags, the industry's relationship to its own tools becomes folkloric — a loop already visible where AI-generated prose quirks are bleeding into human writing, with the metaphor-makers and the models shaping each other's language.
The trend: AI's self-understanding is migrating from technical papers toward folklore, with fiction filling the explanatory gap its own science cannot yet close.