New exam reveals limitations of current AI systems

A new 2,500-question exam, "Humanity's Last Exam," has been developed to test AI capabilities, with current systems failing to achieve high scores. The assessment covers diverse fields like mathematics, humanities, and ancient languages, requiring deep human expertise and context that advanced AI models struggle to replicate, resulting in accuracy below 50%. This initiative aims to provide a more accurate benchmark for AI progress, highlighting the ongoing gap between artificial and human intelligence and the continued importance of specialized human knowledge.