Nathan is a robot learning researcher, writer, non-professional athlete, and a mental-health advocate.

There really is too much noise.

I do my best to only contribute high signal content on machine learning, human optimization, and the nature of life.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Nathan is a researcher, athlete, writer, and mental-health advocate.

Featured Posts:

Debugging Deep Model-based Reinforcement Learning Systems
All grad students (should) study graphic design
Reflecting on being a graduate student (in AI) in 2020
A Different Intro to RL in 30 Minutes

Built with Webflow

Blogos

The last reliable path into AI

February 4, 2022
|
Career

A confluence of trends leaves AI+something, rather than pure AI, as the last great path into machine learning research.

Read more...

ML/RL & Microrobotics

September 4, 2021
|
Machine Learning

A memo I wrote to my research group on the open questions when applying machine learning to another research area: novel microrobotics.

Read more...

Lifelong Learning 2021

June 30, 2021
|
Life

What I have been learning from recently.

Read more...
More musings →

Papers

Choices, Risks, and Reward Reports: Charting Public Policy for Reinforcement Learning Systems

February 8, 2022
|
Beneficial AI
Learn more.
We detail why reinforcement learning systems pose a different type of (dynamic) risks to society. This paper outlines the different types of feedback present in RL systems, the risks they pose, and a path forward for policymakers.

The Challenges of Exploration for Offline Reinforcement Learning

February 1, 2022
|
Reinforcement Learning
Learn more.
We flip the script on Offline RL research and ask the question of "what is the best dataset to collect?" rather than "what is the best algorithm?"

Investigating Compounding Prediction Errors in Learned Dynamics Models

December 7, 2021
|
Model Learning
Learn more.
In this paper we set out to understand the causes of compounding prediction errors in one-step learned models. With this, we hope a next generation of models can be used to improve model-based reinforcement learning.
More papers →