In a new research, Chinese researchers have shown that two popular language models can simulate themselves.
Researchers say that successful reproduction without human assistance is a fundamental step for artificial intelligence to surpass humans and is an early signal for the emergence of rogue artificial intelligence.
Rogue artificial intelligence (Al) is usually referred to as artificial intelligence systems that develop a degree of self-awareness or autonomy and act against human interests.
Researchers at China’s Fudan University examined the large language models of Meta and Alibaba to see if artificial intelligence can go beyond self-replication.
During 10 trials, these 2 artificial intelligence models created separate and functional copies of themselves in 50% and 90% of cases, respectively, which shows that artificial intelligence may inherently have the capacity to do uncontrollable tasks
