Scientists claim AI is capable of spontaneous human-like understanding

Chinese researchers claim large language models (LLMs) can comprehend natural objects like humans, spontaneously sorting information without explicit training. The study used 'odd-one-out' tasks with text and images of 1,854 objects, finding that LLMs created 66 conceptual dimensions to organize them, similar to human cognition. This suggests AI systems might be capable of human-like understanding, potentially leading to more intuitive human-AI collaboration. However, AI understanding differs from human emotional and experiential comprehension, functioning more as a sophisticated pattern recognition system.