I joined watney some time ago. Checkout the only dexterous robots deployed into the real world 24/7.

The distribution of care.

@tonyzzhao This is how a human compares to the trained robot in terms of execution time:

We readapt the ACT implementation by @tonyzzhao that was used to train ALOHA. The policy was trained on an RTX 3080 in <30min. This is how data was collected.

Me and @arbwes are open-sourcing the code to train the low cost robot (<250$) designed by @alexkoch_ai. This is the policy trained on a sorting task with less than 30 demonstrations.

setup

autonomous

In NARRATE we leverage the ability of LLMs to write free-form mathematical expressions such as objective and constraint functions to initialize an optimization-based controller (i.e. MPC) from language instructions. - Project page: - Code (Sim):

How can we empower classical control in robotics with advanced AI *without sacrificing safety and accuracy*?. Introducing NARRATE, a versatile language architecture for optimal control in robotics 🤖. 🧵 Paper, prompts and code open-sourced below 👇

RT @arbwes: I built a web app that generates 3k$ MRR in passive revenue (0 hours a week between me and @ElFrej ). The lesson? It’s not tha….

you should almost always support any startup to succeed. it’s thanks to others’ success that we all get a chance to try it.

My toxic trait is wanting to do a startup in robotics.

RT @sama: How To Be Successful.(At Your Career, Twitter Edition).

RT @ggerganov: Playing some chess using voice. WASM whisper.cpp with a quantized tiny model + grammar sampling (by @ejones). Runs locally i….

RT @arbwes: We made an AI agent a few months ago and now people are using it :).

Lessons are learnt the hard way.