Machines are Learning

on YouTube ↗40m

" summary

Shows concrete claims and tools: training convolutional nets with ~60 million parameters cuts ImageNet top-5 error from ~30% down to ~2.25%, GPUs (EC2 p3 with thousands of tensor cores) speed training from weeks to days, and Stanford models detect diabetic eye disease well enough for FDA-style approval. Demonstrates developer APIs and frameworks—Amazon Rekognition, Amazon Polly, Amazon Lex, TensorFlow/MXNet/PyTorch—builds a working DIY security camera that uploads images, recognizes faces and speaks results in realtime.

AI-written summary · about

" chapters (20)

1. 0:00Start
2. 3:26Unsupervised Learning
3. 3:44Customer Segmentation
4. 4:03Reinforcement Learning
5. 4:40Clustering Analysis
6. 5:23Topic Modeling
7. 6:08Supervised Learning
8. 6:36Binary Classification
9. 8:55Avoid Overfitting
10. 10:02How Can You Avoid Overfitting
11. 11:57Perceptron
12. 13:43Back Propagation
13. 17:03Convolution Matrix
14. 18:56Conclusion
15. 21:43Challenges in Machine Learning
16. 26:14Per-Pixel Object Segmentation
17. 27:43Amazon Recognition
18. 32:55Amazon Lex
19. 36:17Voice to Voice Demo
20. 39:48Questions

" related talks

• 41mDeep Learning to Solve Challenging ProblemsJeff Dean · Google I/O 2019
• 57mBuilding smarter solutions with Machine Learning, from magic to realityLaurent Picard · code.talks commerce 2019
• 47mBuilding smarter apps with Machine Learning, from magic to realityLaurent Picard · GDG DevFest Poland 2018 in Wrocław
• 54mMachine Learning for DevelopersDanilo Poccia · Voxxed Days Zurich 2017
• 01h 03mIt's All About AIJonas Mayer, Martin Förtsch, Thomas Endres · YOW! Melbourne 2022
• 49mKeynote 4: Here Come The Robots - Python and Machine LearningTom Dyson · PyCon APAC 2019
• 59mML and the IoT: Living on the EdgeBrandon D Satrom · NDC Oslo 2019
• 55mMachine Learning and AI in Java on AWSJulien Simon · JavaDay Ukraine 2017