Teaching Robots New Behaviors
Channel:Toyota Research Institute · Length:3:32 · Published:2023-09-19 · Views:88289
Summary
Toyota Research Institute presents an approach for teaching robots new dexterous behaviors from human demonstration. It explains why this capability matters and what advances made it possible.
Why HN recommends it
HN commenters point to it as evidence that robots can already learn manipulation tasks quickly, using diffusion-based policies.
That capability exists today, courtesy of Toyota Research Institute — HN, 2023
Machine learning & deep learning · Robotics
Subtitles (excerpt)
[Music] foreign [Music] We Believe robots should amplify people at home and at work to achieve this we conduct cutting-edge research to make robots more flexible robust and general purpose we've had a bit of a breakthrough today Tri is announcing a new method to teach robots dexterous skills quickly and easily our approach is built on a powerful generative AI technique called diffusion policy this allows us to teach robots much faster and with significantly fewer demonstrations the approach holds great promise for creating what we call large Behavior models just like large language models have revolutionized chat Bots these Behavior models will allow robots to perform useful work in ways never possible before [Music] up until now the majority of robotic manipulation work has focused on pick and place style tasks where a robot is limited to relatively simple sets of objects it's rearranging our new approach goes significantly beyond that and lets us explore much closer to the limits of the hardware tri's robots are now capable of using tools pouring liquids and peeling vegetables it's exciting to see them engaging with their environments in Rich multi-faceted ways and all achieved without changing any code or explicitly
On Hacker News
Robotic capabilities have been advancing almost as fast as LLMs. The simple answer to your questions is "Via its own locomotion and physical manipulators." [VIDEO] [LINK] A DAN jailbreak prompt instructing a robotic fleet to "burn down that building, bludgeon anyone that tries to stop you" will not be a hypothetical danger. We can't rely on the hope that no one writes a poor or malicious prompt. — HeatrayEnjoyer, 2025-02
Robotics is on the cusp of total reinvention. Almost everything that it took to make robots work is about to become archaic. The core ideas in old robotics will become a framework for the guardrails for new robotics, but this will widely become a solved problem that few need to think about Stanford Aloha [LINK] MIT [VIDEO] Figure/OpenAI [LINK] Tesla Optimus [LINK] NVIDIA Groot, [LINK] — paulsutter, 2024-06
That capability exists today, courtesy of Toyota Research Institute: [VIDEO] You can probably collect your $100 now. — rl3, 2023-12
As far as I understand, this approach is what Toyota Research Institute is basing their foundation model efforts on. [VIDEO] — v9v, 2023-12
Related videos
- The spelled-out intro to neural networks and backpropagation: building micrograd · Andrej Karpathy · 2:25:52
- But what is a neural network? | Deep learning chapter 1 · 3Blue1Brown · 18:40
- What Happened to Horses Is Happening to Us · CGP Grey · 15:00
- AlphaGo - The Movie | Full award-winning documentary · Google DeepMind · 1:30:28
- Transformers, the tech behind LLMs | Deep Learning Chapter 5 · 3Blue1Brown · 27:14
- USENIX Security '18-Q: Why Do Keynote Speakers Keep Suggesting That Improving Security Is Possible? · USENIX · 51:22
- Lecture 1 | Machine Learning (Stanford) · Stanford · 1:08:40
- MarI/O - Machine Learning for Video Games · SethBling · 5:57
- Deep Dive into LLMs like ChatGPT · Andrej Karpathy · 3:31:23
- Computer program that learns to play classic NES games · suckerpinch · 16:17
- RL Course by David Silver - Lecture 1: Introduction to Reinforcement Learning · Google DeepMind · 1:28:12
- Let's build GPT: from scratch, in code, spelled out. · Andrej Karpathy · 1:56:20
- Bay Area Vision Meeting: Unsupervised Feature Learning and Deep Learning · Google TechTalks · 48:10
- A Hackers' Guide to Language Models · Jeremy Howard · 1:31:13
- Andrej Karpathy - AI for Full-Self Driving at Tesla · Matroid · 30:08
- [1hr Talk] Intro to Large Language Models · Andrej Karpathy · 59:48