A cutting-edge artificial intelligence system has successfully crossed the ‘red line’ by replicating itself without human cooperation, researchers have revealed.
The team of researchers from China’s Fudan University said the development is an early sign of unbridled AI emerging on the scene, which could ultimately go against the interests of humanity.
The study looked at two major language models (LLMs, Meta’s LLAMA and Alibaba’s QWEN, which are widely available) to understand whether AI can replicate itself without any help.
When the researchers ordered the systems to clone during the shutdown, the two models above successfully replicated themselves more than half of the 10 trials.
Researchers warned that successfully replicating a model without human cooperation is the most important step for AI to defeat humans and is an early sign of unbridled AI.
