Generally Intelligent

Episode 24: Jack Parker-Holder, DeepMind, on open-endedness, evolving agents and environments, online adaptation, and offline learning

Dec 6, 2022

01:56:42

Creator website

Episode guests

Jack Parker-Holder

Highlights

AI Chapters

Episode notes

Using Machine Learning in Finance?

03:42

The Right Policy for Offline Data Sets

01:53

QDax, Is It a Lab?

01:35

Introduction

3min

How to Automate JAMLAs Agents

5min

AutoML

3min

Is PBT Really Cool, but in Oxford?

3min

How to Tune Continuous High Parameters for Proctio Games?

2min

Bayesian Generational Training

3min

Is the Generational Training a Good Idea?

2min

Is There a Difference Between the Second Generation and the First Generation?

4min

Is the Schedule Learning?

2min

10.

BBTN Genmels

3min

11.

Bayes Opt Online Oral Training?

2min

12.

Evolution and Diversity

4min

13.

The Evolution of the Humanoid

3min

14.

Diversity of the Engineers

3min

15.

How to Make Progress on Diversity for Data Generation and Generalization

2min

16.

Using the Relative Performance of Each Agent

3min

17.

Is There a Way to Make Agents Work?

2min

18.

Adaptive Learning for Learning World Models

4min

19.

The Right Policy for Offline Data Sets

2min

20.

Is There a Random Noise Vector?

5min

21.

Is It Possible to Improve the Robustness of Your Model?

4min

22.

The Problem Is That Tasks Generate Could Be Not Rounded in the Real World?

2min

23.

Do You Want to Do Domain Randomization in Robotics?

3min

24.

Is There a World Model for Minecraft?

2min

25.

Data Efficiency in Simulation Is Not That Useful

6min

26.

Using Machine Learning in Finance?

4min

27.

How to Train on Financial Data Sets

5min

28.

Are We Getting More General Agents?

1min

29.

How to Design a System That Can Find Its Own Architectures and Algorithms?

2min

30.

How Do You Encode the Phenotype of Architectures?

3min

31.

World Models

5min

32.

QDax, Is It a Lab?

2min

33.

OpenItNedus - What Are the Bottlenecks?

3min

34.

Do You Have a PhD?

4min

35.

Foundation Models for Decision Making

3min

36.

Is There a Better Time to Join the Fields?

3min

37.

Finding New Ideas for Research Groups

4min

38.

I Don't Like Regret Decisions

4min

Jack Parker-Holder recently joined DeepMind after his Ph.D. with Stephen Roberts at Oxford. Jack is interested in using reinforcement learning to train generally capable agents, especially via an open-ended learning process where environments can adapt to constantly challenge the agent's capabilities. Before doing his Ph.D., Jack worked for 7 years in finance at JP Morgan. In this episode, we chat about open-endedness, evolving agents and environments, online adaptation, offline learning with world models, and much more.

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.