A recent study by the Technion – Israel Institute of Technology has revealed that, unlike artificial intelligence (AI) models that process long texts in their entirety, the human brain creates “summaries” while we read, facilitating the understanding of subsequent information.
Researchers analyzed functional magnetic resonance imaging (fMRI) scans of 219 participants as they listened to stories. They compared brain activity with predictions made by existing AI language models.
They discovered that, while AI models could accurately predict brain activity for short texts (a few dozen words), they failed to do so for longer texts, reports Medical Xpress.
The reason lies in the fact that, while both the human brain and AI models process short texts in parallel (analyzing all words simultaneously), the brain changes strategy for longer texts. Since it cannot process all words at once, it stores a contextual summary—a kind of “knowledge reservoir”—which it uses to interpret subsequent words.
In contrast, AI models process all previously heard text at once, so they do not require this summarization mechanism. This fundamental difference explains why AI struggles to predict human brain activity when processing long texts.
To test their theory, the researchers developed an improved AI model that mimics the brain’s summarization process. Instead of processing the entire text at once, the model created dynamic summaries and used them to interpret future text.
This significantly improved AI predictions of brain activity, supporting the idea that the human brain is constantly summarizing past information to make sense of new input.
This ability allows us to process vast amounts of information over time, whether in a lecture, a book, or a podcast.
Further analysis mapped brain regions involved in both short-term and long-term text processing, highlighting the areas responsible for context accumulation, which enables us to understand ongoing narratives.
This study not only deepens our understanding of how the brain handles linguistic information but also offers valuable insights for developing more efficient AI models that emulate human cognitive processes.
By incorporating summarization mechanisms similar to those of the brain, future AI models could improve in natural language processing tasks and human-computer interaction.
