Generally Intelligent

Episode 18: Oleh Rybkin, UPenn, on exploration and planning with world models

Jul 11, 2022

02:00:40

Creator website

Episode guests

Oleh Rybkin

Highlights

AI Chapters

Episode notes

Morals Enable Latent Learning

05:32

How to Come Up With a Good Idea

02:35

Is There a Benchmark for Adapting to a New Situation?

03:50

Introduction

5min

How Did You Get Into Your Pagete?

4min

Nour Science and the Architecture of the Brain

4min

I've Been Thinking About Reason

5min

Is There a Big Transformer in the Learning Process?

4min

How to Adapt to New Situations?

3min

Do You Have Intuition on How to Get Quick Adoption to New Situations?

1min

Morals Enable Latent Learning

6min

Plan to Explore

5min

10.

The Best Way to Measure Uncertainty Estimation

2min

11.

The Old Radio Isn't Working, Is It?

5min

12.

Theorems of Theoeo Fasti

4min

13.

How to Define Scales?

2min

14.

Using a Robotic Agent

3min

15.

The Agent With Four Legs

3min

16.

The Right Way to Specify Goals

3min

17.

How Do We Learn Language Conditional Agents?

2min

18.

The Problem Is Learning From the Internet

2min

19.

Is There an Alternative to Generalization in Reenforcement Learning?

4min

20.

I Don't Understand Tiktok, I Don't Understand Tick Talk.

3min

21.

Is There a Need for Better Generalization?

4min

22.

How to Train a Moral and Image Nit for a Big Data Set

5min

23.

How to Train a Dynamic Model on Utope?

3min

24.

The Structure People Have Lost to the Giant Transformers in Conpee Mission

6min

25.

Morals Are Important, Right?

4min

26.

Is There a Benchmark for Adapting to a New Situation?

4min

27.

Train Robert to Open a Cabinet in the Google Kitchen

3min

28.

How to Train a Biped Robbert to Walk in Simulation

4min

29.

The Future of Robotic Manipulation

5min

30.

How to Come Up With a Good Idea

3min

31.

The Way People Fail to Write Good Papers

3min

32.

Did a Lot of Things Work?

4min

33.

Plant Explore Does Work, but It's Still Too Simple of a Result

3min

Oleh Rybkin is a Ph.D. student at the University of Pennsylvania and a student researcher at Google. He is advised by Kostas Daniilidis and Sergey Levine. Oleh's research focus is on reinforcement learning, particularly unsupervised and model-based RL in the visual domain. In this episode, we discuss agents that explore and plan (and do yoga), how to learn world models from video, what's missing from current RL research, and much more!

Remember Everything You Learn from Podcasts

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