A NEW AI MODEL AIMS TO MIRROR THE HUMAN MIND

A team of researchers has unveiled Centaur, a foundation model designed to predict and simulate human behavior across a wide range of tasks, marking a bold advance in capturing human cognition.

- 

A STEP TOWARD UNDERSTANDING COGNITION

Built by fine-tuning Meta’s Llama 3.1 70B on a massive dataset called Psych-101, which includes trial-by-trial data from over 60,000 participants making 10 million choices in 160 experiments, Centaur outperforms existing cognitive models.

Lead researcher Marcel Binz announced on X that the model, available on Hugging Face, can simulate human decisions in any experiment described in natural language, from choosing breakfast cereal to tackling complex problems.

Centaur’s strength lies in its ability to generalize to new scenarios, cover stories, and task structures, unlike traditional models limited to specific domains.

It matches or exceeds human performance on benchmarks testing common sense, reasoning, and emotional intelligence, often predicting participant behavior better than the average human.

The model's internal representations also align closely with human neural activity, suggesting it captures key aspects of how the brain processes information. This breakthrough, detailed in an arXiv paper, could reshape cognitive science by offering a tool to study the mind in silico.

The project, backed by 40 researchers, aims to build a unified theory of cognition, a longstanding goal in psychology. While Centaur is a leap forward, it remains a simplification of the human mind, with gaps in modeling emotions or social nuances.

Its development raises questions about how AI might redefine our understanding of thought, with implications for fields from neuroscience to education.

TRAINING ON HUMAN BEHAVIOR