Researchers in China have found that artificial intelligence systems like ChatGPT process information in ways similar to the human brain.
A new study shows that large language models (LLMs), including tools from OpenAI and Google, sort and organise knowledge spontaneously. Surprisingly, they were not trained specifically to do this, making the discovery even more remarkable.
The scientists behind the study come from the Chinese Academy of Sciences and South China University of Technology. They asked ChatGPT-3.5 and Google’s Gemini Pro Vision to perform an “odd-one-out” task, testing how the AI grouped different objects.
In doing so, the AIs created 66 conceptual dimensions to help them understand and categorise what they saw. This behaviour, researchers say, mirrors how humans build mental categories.
They found that these dimensions were not random, but clearly structured and easy to interpret. This suggests the AI is developing knowledge in a way that reflects key aspects of human thought.
The findings challenge the belief that AI only imitates patterns without true understanding. The full study was published in Nature Machine Intelligence.
The researchers believe this work could lead to more human-like AI, helping machines work better alongside people.
Elsewhere, Australian startup Cortical Labs is developing a biological computer that runs on live brain cells. Their system, made from human and mouse neurons, has already learned how to play the game Pong.
The company claims these systems can learn and adapt faster than traditional computers, offering a glimpse into a future where biology and technology combine.
