Episode 19: Minqi Jiang, UCL, on environment and curriculum design for general RL agents

1

Introduction

00:00 • 2min

2

How Did Your Research Interests Develop Over Time?

01:33 • 5min

3

The Overton Window of R L Research

06:19 • 2min

4

Using Self Play in Multi-Agent Learning

07:55 • 3min

5

Random Network Distillation - Is There a Reward?

11:18 • 3min

6

The Learning Potential Score for Revisions of Levels

14:22 • 4min

7

The U C B Bandit and the Generalization Gap?

18:33 • 2min

8

Game Theory

20:42 • 2min

9

Refutation Based Learning Environments

22:15 • 2min

10

Pard Is Optimizing for Minni Max Regret

24:22 • 5min

11

How to Reduce the Reward Reward Expenditure

28:59 • 2min

12

How to Maximize Reputation

30:57 • 3min

13

How to Find the Most Challenging Levels in the Environment

33:54 • 2min

14

Is There a Danger to This?

35:29 • 3min

15

Quality Diversity

38:04 • 3min

16

How to Optimize Towards a Non Stationary Objective?

41:27 • 5min

17

Using a Dreamer Model, Is a Good Idea, Right?

46:14 • 2min

18

Is There a Way to Clone a Poet?

47:49 • 2min

19

Do Hard Behaviours Have a Higher Error Rate?

50:07 • 3min

20

Is There a Difference Between Discovery and Copying Behavior?

52:37 • 2min

21

Language to Second Modality Grounding

54:44 • 2min

22

Learning to Communicate With Machines

56:22 • 3min

23

Is There a Risk of Overfitting?

59:21 • 2min

24

Is There a Zero Shop Prompting Pattern in Multi-Modal GPT Three?

01:01:12 • 3min

25

Getting Used to a Different Day Night Cycle of Life, Right?

01:03:49 • 3min

26

On Line Supervised Learning

01:06:24 • 2min

27

Open Handling

01:08:04 • 4min

28

Open Ended Learning

01:12:28 • 2min

29

Is There a Way to Speed Up Evolution?

01:14:54 • 3min

30

Are You Interested in Deep Learning Architectures?

01:18:17 • 2min

31

Is There a Lot of Innovation in RLS?

01:19:51 • 2min

32

Do You Have Any Controversial Opinions?

01:22:20 • 2min

33

I Think the Rate of Publication Is Way Too High in the Field of Machine Learning

01:24:05 • 3min

34

Is There a Difference Between the Publication Cycle and the Productivity?

01:26:40 • 3min

35

The Difference Between Dreamer Versus Dreamer?

01:29:31 • 4min

36

What Makes a Good Collaboration?

01:33:06 • 2min

37

Is There Any Way to Improve the Performance of Your Agent?

01:34:43 • 2min

38

Adaptive Curricula

01:36:54 • 3min

39

Active Learning as a Curriculum?

01:39:56 • 2min

40

How Did You Get Your Ph.D. In Machine Learning?

01:41:48 • 5min

41

The Future of Machine Learning and Machine Learning

01:46:34 • 2min

42

Unpublished Research Failures

01:48:25 • 2min

43

Is There Something Big Missing?

01:50:55 • 4min