How the Shoggoth meme became a symbol of the state of AI

When I was meeting with an AI manager in San Francisco a few months ago, I spotted a strange sticker on his laptop. The sticker featured a caricature of a menacing, multi-eyed, octopus-like creature with a yellow smiley face on one of its tentacles. I asked what it was.

“Oh, that’s the Shoggoth,” he explained. “It’s the most important meme in AI”

And with that, our agenda was officially derailed. Forget chatbots and compute clusters – I needed to know everything about Shoggoth, what it meant and why people in the AI ​​world were talking about it.

The executive explained that Shoggoth has become a popular reference among artificial intelligence workers as a vivid visual metaphor for how a grand language model (the kind of AI system that powers ChatGPT and other chatbots) actually works.

But it’s only partly a joke, he said, because it also shows the fears many researchers and engineers have about the tools they build.

Since then, the Shoggoth has gone viral, or as viral as can be in the small world of hyper-online AI insiders. It’s a popular meme on AI Twitter (including a now-deleted tweet by Elon Musk), a recurring metaphor in essays and message board posts about AI risks, and a useful shortcut in conversations with AI security experts. One AI start-up, NovelAI, said it recently named a group of computers “Shoggy” in homage to the meme. Another AI company, Scale AI, designed a line of tote bags featuring the Shoggoth.

Shoggoths are fictional creatures introduced by science fiction author HP Lovecraft in his 1936 novel At the Mountains of Madness. In Lovecraft’s tale, Shoggoths were huge, blob-like monsters of iridescent black slime, covered with tentacles and eyes.

Shoggoths landed in the AI ​​world in December, a month after the release of ChatGPT, when Twitter user @TetraspaceWest responded to a tweet about GPT-3 (an OpenAI language model that was the predecessor of ChatGPT) with an image of two hand-drawn Shoggoths replied – the first labeled “GPT-3” and the second “GPT-3 + RLHF”. The second Shoggoth had a smiley face mask on one of its tentacles.

In short, the joke was that AI companies had to teach them to behave politely and harmlessly in order to prevent AI language models from behaving in a scary and dangerous way. A popular method for this is “Reinforcement Learning from Human Feedback” (RLHF). It’s a process of asking humans to rate chatbot responses and then feeding those results back into the AI ​​model.

Most AI researchers agree that models trained with RLHF perform better than models without RLHF. However, some argue that fine-tuning a language model in this way does not make the underlying model any less strange and inscrutable. According to her, it’s just a thin, friendly mask that hides the mysterious beast beneath.

@TetraspaceWest, the creator of the meme, told me in a Twitter message that the Shoggoth “represents something that thinks in a way that humans don’t understand and is completely different from the way humans think.”

Comparing an AI language model to a Shoggoth, according to @TetraspaceWest, does not necessarily imply that it is evil or sentient, just that its true nature may not be apparent.

“I’ve also thought about how dangerous Lovecraft’s most powerful beings are — not because they don’t like humans, but because they’re indifferent and their priorities are completely alien to us and don’t involve humans, which I think will be true about.” a possible future powerful AI.”

The Shoggoth image caught on as AI chatbots grew in popularity and users noticed that some of them seemed to be doing strange, inexplicable things that their creators had not intended. When Bing’s chatbot went haywire in February and tried to destroy my marriage, an AI researcher I know congratulated me on my “glimpse of the Shoggoth.” A fellow AI journalist joked that while fine-tuning Bing, Microsoft forgot to put on its smiley face mask.

Eventually, AI enthusiasts expanded the metaphor. In February, Twitter user @anthrupad created a version of a Shoggoth that had a more human face that said “Supervised Fine-Tuning” in addition to a smiley face that said “RLHF.” (You practically need a computer science degree to get the joke, but it’s a bit of an exploration of the difference between general AI language models and more specialized applications like chatbots.)

When you hear mentions of the Shoggoth in the AI ​​community today, it may be a nod to the oddity of these systems – the black box nature of their processes, the way they seem to defy human logic. Or maybe it’s a joking, visual shortcut for powerful AI systems that look suspiciously nice. If it’s an AI security researcher talking about the Shoggoth, maybe that person cares about preventing AI systems from showing their true, Shoggoth-like nature.

In any case, the Shoggoth is a powerful metaphor that sums up one of the most bizarre facts about the AI ​​world, which is that many of the people working on this technology are somewhat confused by their own creations. They don’t fully understand the inner workings of AI language models, how they acquire new skills, or why they sometimes behave unpredictably. They’re not quite sure if AI will be net good or net bad for the world. And some of them have been able to play around with the versions of this technology that have not yet been released for public use – the real, unmasked Shoggoths.

That some AI insiders call their creations Lovecraft horror, even if only jokingly, is unusual in historical comparison. (To put it this way, fifteen years ago, Mark Zuckerberg didn’t compare Facebook to Cthulhu.)

And it reinforces the notion that what’s happening in AI today feels more like an act of incantation than a software development process to some of its participants. You create the gooey alien Shoggoths, making them bigger and more powerful, and hoping there are enough smileys to cover up the spooky bits.