David Krueger–Coordination, Alignment, Academia

1

Introduction

00:00 • 3min

2

The Consequences of Out Control AI

02:58 • 3min

3

How AI Governance Can Replace SEO

06:03 • 4min

4

The Future of Self Improvement

09:49 • 3min

5

The Importance of Common Sensitivity in Modeling

12:23 • 2min

6

The Importance of Language Models for Safety

14:52 • 3min

7

How to Fine Tune Your Model to Fix a Problem

17:52 • 3min

8

The Importance of Distribution Generalization by Risk

20:25 • 3min

9

The Importance of a Causal Model of the World

23:13 • 2min

10

The Importance of Causality in Machine Learning

24:43 • 4min

11

The Pressure to Build an Agent

28:28 • 4min

12

The Causal Trade in Prisoner's Law

32:02 • 3min

13

The Effect of Gamma on Co-Operation

34:33 • 2min

14

The Importance of Understanding Causality

36:28 • 3min

15

The Limitations of Deep Learning

39:24 • 2min

16

How to Automate Neural Architecture Search

41:51 • 4min

17

The Importance of Scaling Up Deep Learning

45:54 • 2min

18

The Bitter Lesson of Scaling Up Models

48:05 • 2min

19

The Limits of Existential Safety

50:24 • 2min

20

The Evolution of Deep Learning

51:59 • 2min

21

The Importance of Deep Learning in the Field of AI

53:32 • 3min

22

The Importance of Deep Learning in AI

56:08 • 4min

23

The Future of the Information Economy

01:00:32 • 2min

24

The Importance of Existential Safety in Marketing

01:02:24 • 2min

25

The Importance of Coordination in Climate Change

01:04:08 • 2min

26

The Importance of Coordination in AI

01:05:42 • 3min

27

The Role of Coordination in the Future of AI

01:09:10 • 2min

28

The Benefits of Multiple Agents in AI Coordination

01:10:54 • 2min

29

The Importance of Having a Specific Research Agenda

01:12:45 • 5min

30

The Importance of Agendas in Machine Learning

01:17:33 • 2min

31

The Different Intuitions of Researchers

01:19:24 • 2min

32

The Importance of Intuition in AI Alignment

01:20:54 • 2min

33

Reward Modeling for Prosaic Alignment

01:23:00 • 4min

34

The Importance of Incentives in Machine Learning

01:27:23 • 1min

35

The Importance of Intent Alignment

01:28:47 • 2min

36

The Problem With Capitalism

01:30:56 • 2min

37

The Importance of Timelines in Academic Collaboration

01:32:59 • 2min

38

How to Get 50% of the Machine Learning Community to Take This Thing Seriously

01:35:26 • 2min

39

The Importance of Altruism in ML

01:37:12 • 2min

40

The Different Types of Altruism

01:39:22 • 2min

41

The Importance of a Positive Mindset in AI

01:41:22 • 2min

42

The Importance of Alignment in Machine Learning

01:42:53 • 2min

43

How to Spread Awareness of Machine Learning

01:44:31 • 3min

44

The Importance of Coherence in AI

01:47:45 • 2min

45

How to Respond to Arguments About Pretensions

01:49:29 • 3min

46

The Alignment Problem in Foundation Models

01:52:24 • 2min

47

The Future of Machine Learning

01:54:01 • 2min

48

The Evolution of Machine Learning in the Face of the Pandemic

01:56:08 • 4min

49

The Importance of Learning Curves

02:00:21 • 2min

50

The Limits of a Bayesian Model Selection Tool

02:02:05 • 2min

51

The Broken Neural Scaling Law With Zero Breaks

02:04:04 • 2min

52

The Role of Validation in Metadata Archaeology

02:05:38 • 4min

53

How to Train Language Models to Act Assist Humans

02:09:28 • 3min

54

Gold Materialization in Deeper Well With a Low Row

02:12:16 • 4min

55

Keys and Chests: A Natural Experiment

02:16:05 • 4min

56

The Importance of Safety Performance Tradeoffs in Existential Safety

02:19:51 • 5min

57

The Importance of Intuitions in the Design of Smart Systems

02:24:42 • 3min

58

How to Limit Planning to a Short Horizon

02:27:12 • 2min

59

The Importance of Reward Hacking

02:28:54 • 3min

60

How to Optimize a Proxy for Learning

02:31:56 • 3min

61

Reward Hacking in MDPs

02:34:46 • 1min

62

The Importance of Reward Hacking

02:36:15 • 3min

63

The Problem With Reward Functions

02:38:56 • 2min

64

The Evolution of Alignment

02:40:51 • 4min