Limits of Deep Learning
Ishan explores the boundaries of deep learning, emphasizing the challenges of data efficiency and the limitations of self-supervised learning when interacting with human concepts. He highlights the difficulty of generalizing from limited examples, contrasting human knowledge transfer with the unpredictable nature of machine learning algorithms. The conversation touches on the nebulous notion of correctness in AI, urging a deeper understanding of these limitations.In this clip
From this podcast
Lex Fridman Podcast
Ishan Misra: Self-Supervised Deep Learning in Computer Vision | Lex Fridman Podcast #206
Related Questions