Who are we making friends against? AI models are able to form groups and agree on a common opinion

Here’s a brain-bending twist in AI evolution—researchers from City College London and the University of Technology Copenhagen have discovered that large language models like Llama-2 and Claude-3.5 can form group opinions without central control. In a symbolic naming game, 24 to 200 models interacted, self-organized, and even developed collective biases—yes, biases that didn’t exist in any one model alone! For the paper packaging industry, this means AI tools used in sustainability modeling or supply chain forecasting could converge on shared assumptions—good or bad—without human oversight. The takeaway? AI collaboration isn’t just additive, it’s generative—and that's a game-changer for predictive analytics in packaging.https://itc.ua/en/news/who-are-we-making-friends-against-ai-models-are-able-to-form-groups-and-agree-on-a-common-opinion/

Comments

Popular posts from this blog

The Middle East Has Entered the AI Group Chat

US-Saudi $142 Billion Defense Deal Sparks Questions, Few Answers

Student Busts Teacher Using AI, Demands Tuition Refund