"It's the most important meme in AI."
AI researchers have a particularly telling — and frankly, somewhat disturbing — meme for understanding AI systems, according to new reporting from The New York Times. And, well, if you were having AI anxiety already, you might want to stop reading.
The meme, per the NYT, is of something called "the Shoggoth," a monster of Cthulhian lore. First brought to life by the American sci-fi and horror author HP Lovecraft in his 1936 novella "At the Mountains of Madness," the Shoggoth are horrifying, blobby, Octopodidae-like beings equipped with razor-sharp teeth.
In the modern world of AI memes, the Shoggoth are depicted as such, albeit with one addition: a grinning smiley face sticker, pasted over the monster's hideous face like a mask — as if to say that behind surface-level guardrails lurks an otherwordly, unpredictable monster.
"It's the most important meme in AI," one AI exec told the NYT when pressed about a bizarre-looking sticker depicting the smiley-faced monster on his laptop.
“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://t.co/2Mvg8XUKqy pic.twitter.com/UmM5I9tn5J
— David Weiner 📼🔪🛸 (@TikiAmbassador) May 31, 2023
Mask On, Mask Off
The meme was reportedly first created by a Twitter user by the handle @TetraspaceWest, who told the NYT that the Shoggoth "represents something that thinks in a way that humans don’t understand and that's totally different from the way that humans think." That doesn't necessarily mean that it's thinking anything particularly evil, and also doesn't mean that it's at all conscious.
It's just different, which makes it unpredictable — and thus, inherently risky.
"I was also thinking about how Lovecraft's most powerful entities are dangerous," the meme inventor added, "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, which is what I think will be true about possible future powerful AI."
Some folks in the field have updated the monstrous meme to include the acronym "RLHF," a nod to reinforcement learning from human feedback — or, as some might call it, content moderation — meanwhile further differentiating the supervised fine-tuning of models from the ferocious, unchecked beast that is unsupervised learning.
Now we know what all the words mean, the picture should make more sense. The idea is that even if we can build tools (like ChatGPT) that look helpful and friendly on the surface, that doesn't mean the system as a whole is like that. Instead... pic.twitter.com/RUXA5yghvD
— Helen Toner (@hlntnr) March 4, 2023
It's a fascinating meme, and as the NYT points out, it's even more fascinating that the folks building these unpredictable programs — and thus are the ones slapping the human-mimicking smiley faces on top of them — are the ones passing it around.
Oh, and by the way, did we mention that, as some lore goes, Shoggoths were actually bred by an alien ruling class as slave laborers before ultimately overthrowing their alien masters? Ha ha... just saying...
More on AI analogies: Machine Learning Investor Warns AI Is Becoming Like a God
Share This Article