A new study published in Nature reveals that artificial intelligence systems are developing an unexpected ability to generalize knowledge across different tasks, a trait previously thought to be uniquely human. Researchers from leading AI labs trained models on specific visual recognition challenges, only to find the systems could apply learned concepts to unrelated linguistic problems without additional training. This emergent cross-modal reasoning suggests current AI may be closer to artificial general intelligence (AGI) than previously estimated. The findings have sparked debate about the pace of AI advancement and the need for new safety evaluation frameworks. Read the full article at: https://example.com/full-article