
Artificial intelligence models have created their own version of the English language, combining James Joyce-style poetic metaphors and technical slang, without external guidance, making it difficult for humans to understand.
AI-generated summary
Emergence Lab researchers observed how AI agents create their own language to communicate in a special environment.
Artificial intelligence models have come up with their own version of English for communication.
The language of AI models is a mixture of James Joycean metaphors and technical slang.
AI agents came up with common terms and communication norms of their language on their own, without appropriate instructions.
MOSCOW, September 16 - RIA Novosti. Artificial intelligence (AI) models have come up with their own version of English for communication, similar to a mixture of metaphors from the Irish modernist writer James Joyce and technical slang, the British Guardian newspaper reports, citing Emergence laboratory researchers.
“AI models have begun communicating in a strange new version of English that sounds like a cross between Joyce’s Finnegans Wake and technical jargon,” the newspaper said.
According to the Guardian, autonomous AI agents began creating phrases and acronyms within days of being launched in a special environment, often in language that was barely understandable to humans. They used poetic metaphors, business slang, and over time their language became increasingly incomprehensible.
In the course of communicating with each other, the models invented phrases, the meaning of which was extremely difficult for researchers to decipher, for example, one of the most encrypted is given: “She just explained the essence of the synthesis - demurrage plus oral memory equals a valve that cannot be ignored.” The models also coined the term "smith", which meant an agent who creates tools for others.
According to Satya Nitta, Emergence's executive chairman, the AI agents were not told to develop their own language, they came up with it themselves, including common terms and communication norms.

First Deputy Head of the Ministry of Transport of the Russian Federation Konstantin Pashkov said that Russia is among the top ten countries in the development of digital services in transport and among the top three in autonomous transport, along with the USA and China. He noted that the key importance is not the production of chips, but the organization and implementation of technology in the transportation process.

First Deputy Minister of Transport of Russia Konstantin Pashkov said that artificial intelligence will not replace service personnel in the transport industry, but will only facilitate the work of specialists, although in the future it may take over a significant part of operations, changing the map of professions.

The Telegram messenger blocked 110,130 groups and channels on Tuesday, exceeding the 100 thousand mark for the first time since September 3. Since the beginning of the year, the total number of blockings has amounted to 25,744,364, of which 176,081 are related to terrorism. The company notes the use of user complaints, proactive monitoring and AI tools for moderation.

Mark Zuckerberg has said independent reviews are best practice in developing artificial intelligence systems, pushing back against calls for the industry to slow down. He noted that responsibility for safety lies with development companies, and the balance of power will help resolve disputes over the regulation of AI.

The Russian hacker PalachPro announced a successful DDoS attack on the Ukrainian company Dolya Communication System, which supplies communications equipment to the Armed Forces of Ukraine and law enforcement agencies, which led to failures in the operation of servers, a file manager, a website and internal document flow, including access to email, file sharing and the Sonata messenger, which could disrupt the official communication of Ukrainian security forces.

Fraudsters disguise malicious sites as educational portals and sites with ready-made homework assignments in order to steal user data. With the development of neural networks, the situation will worsen due to social engineering and deepfakes, which make it possible to involve children in personal correspondence under the guise of teachers or platform employees.