Researchers at MIT have developed a new AI training method called ‘liquid neural networks’ that allows models to continuously learn from new data without forgetting previously acquired knowledge, a major hurdle known as catastrophic forgetting. The approach mimics the adaptive nature of biological brains, enabling AI systems to update their understanding in real-time, which is crucial for applications like autonomous vehicles and medical diagnostics that encounter evolving information. Initial tests show the liquid networks outperform traditional static models in dynamic environments, though challenges around computational efficiency remain. This advancement represents a significant step toward more flexible and lifelong learning AI systems. Read the full article at https://technologyreview.com/2023/10/ai-liquid-neural-networks.