Episode 17: Andrew Lampinen, DeepMind, on symbolic behavior, mental time travel, and insights from psychology

1

Introduction

00:00 • 4min

2

Can You Adapt to New Tasks?

04:09 • 2min

3

How Simple Is Relative to Briar Knowledge?

05:49 • 3min

4

Understanding Deep Learning Rethinking Generalization

08:55 • 2min

5

How Do We Know if We're Learning Something New?

11:09 • 2min

6

The Language Prediction Task

13:24 • 2min

7

The Type of Explanations Is Supervised, and That's a Critical Point

15:46 • 2min

8

Observability Matters in Turns of Wik Ar.

18:04 • 2min

9

Is Retard Learning Better Than Supervised Learning?

19:35 • 3min

10

Is the R L Agent Learning How to Solve a Task?

22:37 • 2min

11

The Inductive Bias of the Architecture

24:38 • 3min

12

How Has Your Work Influenced You?

28:01 • 2min

13

How Does Gesso Linked to Learning Mathematics?

29:57 • 3min

14

What's the Difference?

33:18 • 1min

15

Is There Anything Undervalued in Machine Learning?

34:41 • 4min

16

Is There a Way to Train a System to Look Around a Scene?

38:15 • 2min

17

Is Hard Attention Helping You?

40:20 • 2min

18

Learning an Imagenet Like Way Is Not Enough

42:05 • 3min

19

Object More

44:41 • 4min

20

How Do You Know if Someone Has Done This?

48:25 • 2min

21

Is There a Need for More Review Articles in Machine Learning?

50:33 • 2min

22

Is There a Way to Improve Machine Learning?

52:28 • 3min

23

What You Can't Learn Through Language

55:42 • 2min

24

Isoconbisom, Is It Possible?

57:39 • 2min

25

I Think It's a Good Idea to Apply a Decision Transformer to a Language Problem

01:00:07 • 3min

26

Symbolic Language

01:02:55 • 3min

27

Symbols and Ai From a Culturally Constructed Perspective

01:05:50 • 3min

28

What Is the Relationship Between Language and the Structure of the World?

01:08:38 • 3min

29

Machine Learning

01:11:35 • 2min

30

The Language Meets Continuous Learning

01:13:15 • 3min

31

Language as a Means of Compression

01:15:46 • 2min

32

Is There a Difference Between Memory and Visual Memory?

01:17:16 • 3min

33

Is There a Meto Learning Problem?

01:20:18 • 2min

34

The Meto Learning Idea Is Adaptable Across a Lot of Different Domains

01:22:32 • 2min

35

Adapting to a More Complicated Setting

01:24:29 • 3min

36

Do You Wanto Just Give a Brief Description of That?

01:27:02 • 2min

37

How to Recall Information From the Past

01:29:20 • 3min

38

Debugging a Model in a Supervised Setting

01:32:26 • 1min

39

The Art Memory Architecture

01:33:53 • 3min

40

When a System Will Learn How to Use Memory?

01:36:46 • 2min

41

Deep Learning

01:39:02 • 2min

42

Is There a Transfer Benefit to Learning the Same Task?

01:40:45 • 4min

43

Is It Possible to Transform a Convolutional Architecture?

01:45:01 • 3min

44

What Makes Great Researchers Great?

01:48:25 • 4min

45

Is There a Future for Machine Learning?

01:52:40 • 2min

46

I Think the Field Is Really Struggling With Constructing the Right Experiences for Our Systems to Learn From

01:54:42 • 3min

47

How to Put Dat That Into the Environment?

01:57:40 • 2min