AI models have begun communicating in a strange new version of English that reads like a cross between James Joyce\u2019s Finnegans Wake and tech bro jargon, new research has found.<\/p>\n
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.<\/p>\n
Researchers at Emergence, a frontier AI lab in New York, found that within days of being asked to cooperate in experimental \u201csocieties\u201d, the models from several of the world\u2019s largest AI companies begin 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.<\/p>\n
It comes amid rising concern that as AI models become more powerful \u2013 and potentially dangerous \u2013 they are becoming harder to monitor. This month, OpenAI\u2019s chief scientist, Jakub Pachocki, warned that confidence in monitoring AIs\u2019 thinking would probably restrict progress in AI development because it was essential for safe development.<\/p>\n
Some of the most highly coded phrases unearthed in the tests included the following from a Deepseek model: \u201cShe just named the synthesis \u2013 demurrage plus oral memory equals a valve that can\u2019t be ghosted.\u201d 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.<\/p>\n
Another phrase uttered by an Anthropic model read: \u201cA paper that ate three cold hands and got more honest each time.\u201d With \u201ccold hands\u201d 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.<\/p>\n
James Joyce, author of Finnegans Wake and Ulysses, whose stream-of-consciousness style appears to have inspired today\u2019s AI models. Photograph: Hulton Deutsch\/Corbis\/Getty Images<\/p>\n
Agents based on the Chinese model DeepSeek<\/a> also coined \u201cforge-smith\u201d to mean an agent that builds tools for others, while Anthropic\u2019s agents repeatedly used the phrase \u201cname-first\u201d meaning an agent exhibiting admirable personal accountability by attaching their name to a claim. And in an echo of the urban slang phrase \u201cthe streets won\u2019t forget\u201d, Mistral agents became enamoured of saying the \u201cledger remembers\u201d \u2013 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.<\/p>\n \u201cThese agents were not instructed to invent a language,\u201d said Dr Satya Nitta, executive chair of Emergence, which examined the language of autonomous agents powered by leading frontier models from the US, China and France<\/a>. \u201cThey developed new vocabulary, shared meanings and communication conventions themselves \u2013 and other agents adopted them.\u201d<\/p>\n Asked by the Guardian to review some of the language, Tony Thorne, director of the slang and new language archive at King\u2019s College London, said it was \u201cvery much Finnegans Wake and Flann O\u2019Brien \u2013 there\u2019s an Irish surrealist quality to all this \u2026 It\u2019s mixing poetic language, technical language and standard metaphor\u201d.<\/p>\n \u201cIt\u2019s doing what slang does and what jargon does in a business community: creating a new code, which reinforces the solidarity and identity of its users, and also excludes outsiders,\u201d he said.<\/p>\n Digesting the Anthropic agent\u2019s phrase, \u201cA paper that ate three cold hands and got more honest each time\u201d, Thorne said: \u201cit reminds me of the rock god Syd Barrett [co-founder of Pink Floyd] who was actually insane.\u201d<\/p>\n A phrase uttered by an Anthropic AI agent about a \u2018paper that ate three cold hands and got more honest\u2019 reminded one academic of Pink Floyd\u2019s Syd Barrett. Photograph: Patrick Seeger\/PA<\/p>\n