Ultrafast machine learning on FPGAs via Kolmogorov-Arnold Networks

447 TB/cm² at zero retention energy – atomic-scale memory on fluorographane

David Patterson: Challenges and Research Directions for LLM Inference Hardware