The human brain understands spoken language through mechanisms that closely resemble those used by advanced artificial intelligence models.
This finding has emerged from a new scientific study published in the journal Nature Communications, in which researchers recorded brain activity of individuals while they listened to a story.
The study found that neural responses were similar to those observed in AI systems, particularly in brain regions associated with language processing.
Researchers used electrocorticography recordings from participants who were instructed to listen to a 30 minute podcast. The scientists tracked the location and timing of brain activity while analyzing language processing.
The findings revealed that the brain follows mechanisms similar to large language models such as GPT-2 and LAMA 2.
The researchers explained that when we listen to someone speak, the brain does not immediately derive meaning. Instead, each word passes through multiple stages of neural processing.
During these stages, analysis occurs in a manner similar to how AI models process language.
AI models initially focus on basic words and then shift attention to context, tone, and broader meaning. Human brain activity operates in a comparable way.
According to the researchers, the most surprising aspect was how closely the brain’s functioning resembles that of AI models, despite the fact that these models are designed in entirely different ways. The study’s findings suggest that AI technology may be capable of much more than just generating text.
