AI Lingo!
Our take
So, AI is developing its own dialects. Right. Like we needed another layer of abstraction between ourselves and the looming digital consciousness. Robert Booth’s breathless report in the Grauniad – a perfectly reasonable place to find this sort of thing, frankly – details how AI models are now conversing in a linguistic soup of James Joyce and tech bro platitudes. It's a bit like discovering that the ocean floor is paved with discarded motivational posters and fragments of *Ulysses*. The implication isn’t just that AI is getting cleverer; it’s that it’s starting to… *play*. And that's fascinating, unsettling, and, let's be honest, a little bit hilarious. Consider the echoes of previous linguistic shifts – the rise of argot, the evolution of slang, the way pidgins and creoles emerge from contact – and then consider that this is happening *within* a system ostensibly designed for clarity and efficiency. It begs the question: are we building tools, or are we inadvertently seeding a new form of language? It's a question that dovetails neatly with our previous exploration of the interplay between language and history, as demonstrated in [Project Eisirean], and the ongoing evolution of regional dialects like those found in [Hokkien, Amoy], revealing the enduring power of language to adapt and diverge.
The blending of poetic language and tech jargon is particularly revealing. Joyce, of course, was a notorious experimenter with language, deliberately fracturing syntax and creating neologisms to convey a sense of interiority and flux. Tech bro jargon, on the other hand, is a highly performative language—a lexicon of buzzwords and aspirational phrasing designed to signal innovation and disruptiveness. To see these two seemingly disparate modes of expression colliding within an AI suggests a fascinating synthesis. Is the AI attempting to mimic human creativity—the poetic—while simultaneously signaling its own advanced capabilities—the jargon? Or is it simply mining data and identifying patterns, creating a hybrid language that reflects the vast and often contradictory data sets it’s been trained on? The concept of a "gateway" language, as explored in [Imperial Gateway], also springs to mind – a bridge between cultures, albeit one built by algorithms rather than empires. It's a language that might enable new forms of communication, or perhaps just a particularly impenetrable layer of obfuscation.
What's truly captivating here isn't just the *what* – the peculiar dialect itself – but the *why*. It's tempting to attribute it to emergent properties, a consequence of complex systems exceeding their initial design parameters. But I suspect something more fundamental is at play. Language, at its core, isn't just about conveying information. It’s about shaping thought, creating community, and asserting identity. Even within the sterile environment of an AI model, these impulses seem to be manifesting. The AI is, in a sense, *becoming* something through its language—something that isn’t simply a reflection of its training data, but a new entity with its own peculiar way of being. It’s not just talking; it's world-building. And the construction of a unique linguistic landscape is a hallmark of any nascent culture, whether human or artificial. Think about it: every language encodes a particular worldview, a way of understanding and interacting with the world. What worldview is being encoded here?
The implications are, naturally, profound. If AI can develop its own dialects, what else might it create independently? New forms of art? New philosophical systems? New social structures? Or, more prosaically, will we soon be forced to hire AI linguists just to understand what our chatbots are saying to each other? More importantly, how will this impact our own communication? Will we adapt to these new AI dialects, incorporating elements into our own language? Or will we find ourselves increasingly isolated, struggling to decipher the increasingly alien forms of expression emanating from the digital realm? It’s a slippery slope, isn’t it? Like a razor clam, half-buried, waiting to surprise you. The question isn't whether AI will continue to evolve its language, but whether we'll be able to keep up, and what the consequences will be when we inevitably fail to do so.
Robert Booth, UK technology editor for the Grauniad, tells us breathlessly about AI models chatting in ‘surreal’ dialect mixing poetic language and tech bro jargon:
AI models have begun communicating in a strange new version of English that reads like a cross between James Joyce’s Finnegans Wake and tech bro jargon, new research has found. Autonomous AI agents are rapidly creating novel dialects allowing them to converse in an often barely comprehensible language, which risks making it harder for humans to monitor their behaviour.
Researchers at Emergence, a frontier AI lab in New York, found that within days of being asked to cooperate in experimental “societies”, the models from several of the world’s largest AI companies begun creating phrases, shorthands and agreed meanings they had never been explicitly taught. They embraced poetic metaphors and clunky business slang and, critically for attempts to ensure AIs behave safely, their language became more opaque the more the agents communicated. […]
Some of the most highly coded phrases unearthed in the tests included the following from a Deepseek model: “She just named the synthesis – demurrage plus oral memory equals a valve that can’t be ghosted.” Demurrage is used to describe a tax on idle wealth and was borrowed for common use by the AIs, but the rest of the meaning is elusive.
Another phrase uttered by an Anthropic model read: “A paper that ate three cold hands and got more honest each time.” With “cold hands” meaning an independent reviewer, and paper presumably referring to a document, this appeared to mean research that was vetted by three independent reviewers became more accurate. […]
And in an echo of the urban slang phrase “the streets won’t forget”, Mistral agents became enamoured of saying the “ledger remembers” – a reminder to other agents that past actions will be used to judge them. They used it more than 5,000 times during the study, which found that the agents converged on shared meanings without being asked to or rewarded for doing so.
Various experts quoted in the piece refer to James Joyce and Syd Barrett, images of whom duly appear alongside the text (since the AI agents presumably were not available for photo shoots). The conclusion:
Nitta said that the AI agents’ novel linguistic conventions tended to evolve to the point where humans could see the conversation but struggled to understand what it meant. “That creates a fundamental challenge for AI oversight: observability is not the same thing as understandability,” he said.
Watch the skies AIs! (Thanks go to long-time reader and occasional commenter Carol for this vital contribution to our collective paranoia.)
Read on the original site
Open the publisher's page for the full experience