Explore a comprehensive analysis of self-supervised learning in artificial intelligence through this 59-minute video lecture. Delve into the concepts of supervised learning, self-supervised learning, and common sense in AI systems. Examine the process of predicting hidden parts from observed parts and compare self-supervised learning applications in language and vision. Investigate energy-based models, joint-embedding models, and contrastive methods. Learn about latent-variable predictive models and GANs. Gain insights from Yann LeCun and Ishan Misra's research at Facebook AI, discussing the potential of self-supervised learning as a key approach to developing AI systems with improved background knowledge and common sense capabilities.
Self-Supervised Learning - The Dark Matter of Intelligence
Overview
Syllabus
- Intro & Overview
- Supervised Learning, Self-Supervised Learning, and Common Sense
- Predicting Hidden Parts from Observed Parts
- Self-Supervised Learning for Language vs Vision
- Energy-Based Models
- Joint-Embedding Models
- Contrastive Methods
- Latent-Variable Predictive Models and GANs
- Summary & Conclusion
Taught by
Yannic Kilcher
Related Courses
-
Energy Based Models and Self-Supervised Learning
-
Latent Variable EBMs for Structured Prediction
-
Contrastive Methods and Regularised Latent Variable Models
-
AI and Machine Learning Algorithms and Techniques
-
A Path Towards Autonomous Machine Intelligence - Paper Explained
-
Joint Embedding Method and Latent Variable Energy Based Models
