Challenges & future trends of Deep Learning
summary
Autonomous cars generate data equal to 2,600 average internet users and that scale drives both opportunity and cost; a real-time mask detector and an AOI defect inspector are built and run successfully using YOLOv3, training in a few hours on one GPU with only a few hundred images. GANs produce photorealistic deepfakes in minutes via tools like avatarify. Core claims: current systems are supervised and brittle, the next step is disentanglement learning and unsupervised, robust lifelong models.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 0:45Outline
- 1:40Trinity of AI
- 2:55Data
- 3:42Autonomous cars
- 4:25Realtime object detection
- 5:28Gans
- 6:40Deep Fake
- 8:55Deepfake videos
- 9:30Frontier Ni
- 12:07Ultimate Destination
- 13:38Automatic Optical Inspection
- 13:50Cosmetic Defects
- 15:35Machine Learning Computer Vision
- 17:22YOLO Example
- 18:16GPU Training
- 18:40Python and R
- 19:26Python vs R
- 19:57Best Program Tools
- 20:17Python
- 20:44Conclusion
- 22:24Questions
related talks
42mFrom theory to practice: The rise of Generative AI
01h 03mIt's All About AI
13mNavigating Generative AI and Its Risks
41mThe AI boom: what and why?
56mDeep Fakes 2.0 - How Neural Networks are Changing our World
31mDeep learning for self-driving vehicle technology
41mDeep Learning to Solve Challenging Problems
29mDeep Learning, Neuroscience and the future of AI