A recent essay highlighted by The Decoder claims that approximately 90 percent of neurons in neural networks serve a distinct and identifiable function. This challenges the common belief that AI systems are largely opaque and difficult to interpret.
The finding has sparked renewed debate among researchers about AI interpretability and the potential for better understanding how neural networks operate internally. If neurons can be more easily mapped to specific tasks, it could improve transparency and trust in AI models.
For Japanese markets, where AI adoption in sectors like finance and technology continues to grow, clearer insights into neural network behavior may enhance risk assessment and regulatory compliance.