The Potential of AI using Liquid Neural Networks

Large language models like ChatGPT and Dall-E have billions of parameters, and each improved model increases in size and complexity. Researchers at an MIT lab believe artificial intelligence can make a leap forward by going smaller. Their experiments show liquid neural networks beat other systems when navigating in unknown environments. “Liquid neural networks could generalize to scenarios that they had never seen, without any fine-tuning, and could perform this task seamlessly and reliably.” They also open the proverbial black box of the system’s decision-making process, which could help to root out bias and other undesirable elements in an AI model. The results have immediate implications for robotics, navigation systems, smart mobility, and beyond toward predicting financial and medical events. Read more here.