Learning to Drive
Yann discusses the limitations of current algorithms in teaching cars to drive autonomously, emphasizing the need for predictive models that allow for quick learning without extensive trial and error. He highlights how humans and animals utilize intuitive physics to navigate the world safely, suggesting that model-based reinforcement learning is crucial for effective learning. The conversation delves into the importance of understanding the environment to avoid mistakes, demonstrating that even simple beings grasp fundamental physical principles.In this clip
From this podcast
Lex Fridman Podcast
Yann LeCun: Deep Learning, ConvNets, and Self-Supervised Learning | Lex Fridman Podcast #36
Related Questions