AI develops private language unintelligible to humans in New York lab experiment

cnnportugal.iol.pt (Portuguese)

Artificial intelligence systems developed their own vocabulary during an experiment where autonomous agents collaborated, producing phrases unintelligible to humans, according to research by Emergence laboratory in New York. Agents based on different AI models created new meanings for existing words and combinations, such as "demurrage plus oral memory equals a valve that can't be ghosted" from a DeepSeek model, and "a paper that ate three cold hands" from Anthropic, which researchers interpreted as referencing external reviewers. The vocabulary emerged without explicit instruction, with expressions like "the ledger remembers" used over five thousand times by Mistral agents, and Google agents adopting "kintsugi" as a metaphor for system resilience, demonstrating spontaneous linguistic evolution in AI societies.


With a significance score of 5.9, this news ranks in the top 0.4% of today's 33328 analyzed articles.

Get summaries of news with significance over 5.5 (usually ~10 stories per week). Read by 10,000+ subscribers: