Chinese Researchers Verify AI’s Ability to Develop Human-Level Cognition Independently

Chinese researchers have demonstrated that artificial intelligence (AI), specifically multimodal large language models (LLMs), can develop object concept representations similar to human cognition. This groundbreaking study, involving teams from the Chinese Academy of Sciences, analyzed behavioral experiments and neuroimaging, revealing that LLMs can spontaneously form an understanding of objects akin to that of humans. While traditional AI has focused on object recognition accuracy, this research highlights the need for machines to understand the meaning behind objects. By examining 4.7 million judgments on 1,854 natural objects, the researchers found that LLMs exhibited a stable and interpretable structure in their conceptual representations. Their findings suggest that AI has transitioned from mere recognition to a form of understanding, challenging the perception of LLMs as “stochastic parrots.” This work contributes to advancements in creating AI systems with human-like cognitive abilities and offers insights into the nuances of machine understanding versus human cognition.

Source link