AI Pioneer on whether it's time to stop celebrating new LLMs until hallucination is fixed | HTLS 2026

Can AI become an independent scientific collaborator, or is the promise of autonomous AI scientists overhyped? At the 24th edition of the Hindustan Times Leadership Summit, Professor Anima Anandkumar, Bren Professor at Caltech, explains why the biggest challenge in scientific discovery is not generating new ideas, but verifying them through accurate simulations and physical experiments. Anandkumar discusses how AI-powered digital twins could reduce the cost and time involved in testing new materials and scientific discoveries. She also addresses concerns around AI hallucinations, explains how physics-informed neural networks can deliver high-precision results, and highlights the importance of mathematical verification. She argues that AI must understand the physical world and follow the laws of physics if it is to truly accelerate scientific breakthroughs.