AI Weekly Update - June 16th, 2021 (#35!)

Content Links Below:

Chapters
0:00 Introduction
15:17 Generative Models as a Data Source for Multi-View Representation Learning
18:14 Learning to See by Looking at Noise
21:01 Knowledge Distillation: A Good Teacher is Patient and Consistent
24:30 Does Knowledge Distillation Really Work?
26:37 AdaMatch
29:47 Self-Damaging Contrastive Learning
31:35 Masked Self-Supervised Transformer for Visual Representation
33:12 Data-Efficient Instance Generation from Instance Discrimination
34:34 Scaling Vision Transformers
36:48 CoAtNet
38:17 Improving Language Model Behavior by Training on a Curated Dataset
39:11 Dynamic Language Models for Continuously Evolving Content
40:38 GPT-J-6B
41:43 An Empirical Survey of Data Augmentation for Limited Data Learning in NLP
43:03 Meta-Learning with Fewer Tasks through Task Interpolation
44:24 Exploring the Limits of Out-of-Distribution Detection
45:10 Causality in Deep Learning
46:42 A graph placement methodology for fast chip design
47:49 Pretraining Representations for Data-Efficient Reinforcement Learning
49:20 CodeXGLUE
49:52 HF Zero Shot Classification
50:24 HuggingFace Course
50:52 Neural Structured Learning
51:09 GPT-3 vs. GPT-Neo Data Picture
52:12 Audio to video generation
52:55 New PaperswithCode Datasets

[um]

Thank you very much for doing these types of videos! The series is really helpful and provides a snapshot of interesting and valuable works in what has become an over hyped field riddled with projects presented as research.

[um]

Thank you very much for doing this series! It's a fantastic snapshot of interesting and valuable works in what has become an over hyped field riddled with projects presenting as research.

[roughr]

It's really helpful to get some of latest advancements in DL in a single video... Thank you!

[juanmanuelcirotorres]

First, thanks, as always very valuable information. Only one comment, is there a reason to don't divide the video into smaller ones? I think that is in that way you can get more views and is a little easier see it, regards from Colombia