AI may be able to think like humans after all
What's the story
A group of Chinese researchers has provided the first evidence that artificial intelligence (AI) large language models can spontaneously form a human-like system to understand and categorize natural objects.
This process is considered a fundamental aspect of human cognition.
The study was published in the peer-reviewed journal Nature Machine Intelligence earlier this week.
Research focus
Can LLMs develop human-like object representations?
The research team explored whether LLMs could develop human-like object representations from linguistic and multimodal data. LLMs are AI models trained on massive amounts of text, as well as visual and audio data in the case of multimodal large language models (MLLMs).
Insight
Debate over cognitive capacity of AI models
The study also emphasized the importance of understanding how humans conceptualize and categorize natural objects.
It said, "Understanding how humans conceptualize and categorize natural objects offers critical insights into perception and cognition."
This research provides new evidence in the ongoing debate over the cognitive capacity of AI models, indicating that artificial systems that reflect key aspects of human thinking may be possible.
Steps
How was the study done?
The research was conducted by scientists from the Institute of Automation at Chinese Academy of Sciences (CAS) as well as South China University of Technology.
The scientists used ChatGPT-3.5 to perform tasks using solely text-based descriptions, while Gemini Pro Vision 1.0 was utilized to perform parts of the study using pictures.