Recent research from the University of California, Berkeley, suggests that artificial intelligence (AI) systems can process signals in a similar way to how the brain interprets speech. This finding has significant implications for understanding how AI systems operate.
Researchers from the Berkeley Speech and Computation Lab placed electrodes on the heads of participants while they listened to a single syllable, “bah.” They then compared the resulting brain activity to the signals produced by an AI system trained to learn English.
The study found that the Artificial Intelligence system’s signals closely mirrored the brain’s processing of speech, particularly in the convolutional layers of the neural network and the brainstem. Interestingly, the study also found that how speech is encoded in these regions varies based on an individual’s language experience.
Gasper Begus, assistant professor of linguistics at UC Berkeley and lead author on the study, said, “The shapes are remarkably similar. That tells you similar things get encoded, that processing is similar.”
A side-by-side comparison graph of the two signals shows that similarity strikingly. Begus added, “There are no tweaks to the data. This is raw.”
This research sheds light on the “black box” of AI systems, which can be difficult to understand due to their complex algorithms. Understanding how AI systems process signals in a way similar to the brain may help researchers improve the performance of these systems in the future. Additionally, the findings may have implications for the development of more advanced neural interfaces, which could allow for direct communication between the brain and AI systems.
The Berkeley Speech and Computation Lab study revealed that AI systems could process speech signals similar to the brain. The study analyzed the signals produced by an AI system and compared them to brain activity in regions involved in speech processing.
This discovery may aid researchers in developing better-performing AI systems and neural interfaces, allowing for direct communication between the brain and AI systems.
The study conducted by the Berkeley Speech and Computation Lab used electrodes to measure brain activity while participants listened to a syllable. The study then compared brain activity to signals produced by an AI system trained to learn English.
The researchers analyzed the signals produced by the convolutional layers and brainstem, regions involved in speech processing.
In conclusion, the study conducted by researchers at the Berkeley Speech and Computation Lab showed that AI systems process speech signals in a way that closely mirrors the brain’s processing of speech, particularly in the convolutional layers of the neural network and the brainstem. The new study also found that language experience can influence how speech is encoded in these regions.
These findings provide important insights into the similarities between the neural networks of the brain and AI systems. They may have implications for the development of more advanced neural interfaces and the future development of AI systems.
Journal Reference:
- Beguš, G., Zhou, A. & Zhao, T.C. et al. Encoding of speech in convolutional layers and the brain stem based on language experience. Scientific Reports. DOI: 10.1038/s41598-023-33384-9