AI language models develop social norms like groups of people

AI language models develop social norms like groups of people

When LLMs are grouped together, they exhibit similar characteristics to human societies. When LLMs are grouped together, they exhibit similar characteristics to human societies.

Artificial intelligences can dialogue with each other and change their language in relation to this dialogue. Groups of artificial intelligence agents based on large language models (LLM) can spontaneously organize themselves in society, developing shared social conventions without any direct human intervention. This is revealed by a study conducted by researchers from City St Georges, University of London and the IT University of Copenhagen, pu…

LLMs Struggle with Real Conversations: Microsoft and Salesforce Researchers Reveal a 39% Performance Drop in Multi-Turn Underspecified Tasks

Conversational artificial intelligence is centered on enabling large language models (LLMs) to engage in dynamic interactions where user needs are revealed progressively. These systems are widely deployed in tools that assist with coding, writing, and research by interpreting and responding to natural language instructions. The aspiration is for these models to flexibly adjust to changing user inputs over multiple turns, adapting their understan…

Can these models coordinate their behavior by forming conventions, building blocks of a society?

What happens if AIs like ChatGPT don't talk to people, but only to each other? A new study shows that they form their own standards – without rules, planning or central control. read more on t3n.de

AI agents can learn human social norms, study reveals

Artificial Intelligence (AI) technology can adopt human-like social behaviors, according to a new study. This finding was revealed in research conducted by St. George’s University of London and the IT University of Copenhagen. The study indicated that AI agents based on Large Language Models (LLMs) such as ChatGPT are capable of interacting in group conversations without any external interference. These agents can adopt linguistic features and s…