Israeli and U.S. researchers have made a significant discovery regarding language processing in the human brain. They found that the brain processes spoken language in a sequential manner that closely reflects the internal mechanisms of advanced artificial intelligence, as stated by the Hebrew University of Jerusalem.

The findings, published in the journal Nature Communications, indicate a remarkable similarity in how the human brain and Large Language Models (LLMs) derive meaning, despite their different structural foundations.

In their research, scientists from the Hebrew University of Jerusalem, alongside colleagues from Princeton University and Google Research, observed that when the brain listens to speech, it translates words into meaning through a rapid sequence of neural steps. This unfolding process occurs over time and correlates with how AI models process information through multiple layers.

The study demonstrated that early brain responses to speech match the initial surface layers of an AI model, which focus on basic linguistic features. Following this, brain activity aligns with the deeper layers of AI models, where context, tone, and intricate meanings are integrated.

This notable alignment was particularly observed in Broca's area, the primary language center of the brain, where the most robust activity corresponded to the most advanced layers of AI models.

The discovery fundamentally challenges traditional linguistic theories that view language processing through a strict, rule-based lens. Instead, it supports a model where meaning gradually emerges through context, mirroring the operation of AI.

To facilitate further exploration into the intricacies of how the brain deciphers natural language, the research team has made their complete dataset of brain recordings and language features publicly accessible.