Democratizing Automation
Subscribe
Sign in
Home
Chat
Archive
About
New
Top
Discussion
Scaling laws for robotics & RL: Not quite yet
Robotics Transformers, DreamerV3, XLand 2, and hoping that scaling laws are coming embodied AI.
Nathan Lambert
Feb 1
5
4
Share this post
Scaling laws for robotics & RL: Not quite yet
robotic.substack.com
Copy link
Twitter
Facebook
Email
January 2023
Pretraining quadrupeds: a case study in RL as an engineering tool
How an unlikely corner of robotics research, locomotion, defined RL's new notion of success.
Nathan Lambert
Jan 16
5
3
Share this post
Pretraining quadrupeds: a case study in RL as an engineering tool
robotic.substack.com
Copy link
Twitter
Facebook
Email
Looking into 2023
My predictions for machine learning this year: 3D assets, self-driving, GPT4, RLHF, Deep RL, diffusion models, and conference cycles.
Nathan Lambert
Jan 6
5
4
Share this post
Looking into 2023
robotic.substack.com
Copy link
Twitter
Facebook
Email
December 2022
Predicting machine learning moats
Models aren't moats and how emergent behavior scaling laws will change the business landscape.
Nathan Lambert
Dec 28, 2022
20
6
Share this post
Predicting machine learning moats
robotic.substack.com
Copy link
Twitter
Facebook
Email
Closed-API vs Open-source continues: RLHF, ChatGPT, data moats
Model-as-a-service makes more sense when there is a data advantage to back it up.
Nathan Lambert
Dec 19, 2022
12
3
Share this post
Closed-API vs Open-source continues: RLHF, ChatGPT, data moats
robotic.substack.com
Copy link
Twitter
Facebook
Email
RLHF, 'online' ML systems, and RL going mainstream
Common machine learning systems are starting to deploy the RL lens of feedback.
Nathan Lambert
Dec 5, 2022
7
5
Share this post
RLHF, 'online' ML systems, and RL going mainstream
robotic.substack.com
Copy link
Twitter
Facebook
Email
November 2022
Join my new subscriber chat
A private space for us to converse and connect
Nathan Lambert
Nov 10, 2022
2
2
Share this post
Join my new subscriber chat
robotic.substack.com
Copy link
Twitter
Facebook
Email
October 2022
Using RL's exploitation to debug
A musing on how I think autonomous system companies should use RL.
Nathan Lambert
Oct 26, 2022
4
2
Share this post
Using RL's exploitation to debug
robotic.substack.com
Copy link
Twitter
Facebook
Email
September 2022
Back in the game
What I've been up to and what's coming soon.
Nathan Lambert
Sep 26, 2022
4
Share this post
Back in the game
robotic.substack.com
Copy link
Twitter
Facebook
Email
February 2022
Designing Societally Beneficial Reinforcement Learning Systems
Choices, risks, and reward reporting. Recommendations for how to integrate RL systems with society.
Nathan Lambert
Feb 8, 2022
1
Share this post
Designing Societally Beneficial Reinforcement Learning Systems
robotic.substack.com
Copy link
Twitter
Facebook
Email
January 2022
Flexible Centralization in Multi-agent Learning & Control
A tour of control theory, multi-agent RL, and hierarchical learning.
Nathan Lambert
Jan 21, 2022
2
Share this post
Flexible Centralization in Multi-agent Learning & Control
robotic.substack.com
Copy link
Twitter
Facebook
Email
August 2021
Remote robotic-data farms
Industry labs power up their robot learning research with parallelization!
Nathan Lambert
Aug 9, 2021
2
Share this post
Remote robotic-data farms
robotic.substack.com
Copy link
Twitter
Facebook
Email
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts