Subhashis Banerjee

CNNs struggle with unseen concepts, while LLMs show surprising clustering of related topics, challenging our understanding of internal representations in neural architectures.

Validity of AI inferences should be a non-negotiable condition for deployment, not just an afterthought in regulatory frameworks.