Chinese researchers say AI models mimic human-like understanding

(ECNS) -- Chinese researchers have confirmed for the first time that AI models can form human-like conceptual understanding - a breakthrough that could reshape how scientists think about machine cognition, the Chinese Academy of Sciences said Tuesday.

Drawing on principles of human cognitive science, researchers gave AI models and human participants three object concepts randomly selected from a pool of 1,854 common items and asked them to identify the one that didn’t belong. By analyzing 4.7 million such comparisons, they created the first-ever “thought diagram” mapping how the model processes abstract concepts.

He Huiguang, corresponding author of the paper and a researcher at the CAS Institute of Automation said researchers identified 66 key dimensions that explain how the model interprets objects - many of which closely mirror brain activity in humans.

Even more significantly, the multimodal model, capable of interpreting both text and imagery, was found to reason and make decisions in a way more closely aligned with human thinking than other models.

The findings were published online in the journal Nature Machine Intelligence. Researchers said these findings help advance the understanding of machine intelligence and inform the development of more human-like artificial cognitive systems.