55 years after being part of “Ping-Pong Diplomacy”, a U.S. table tennis player returns to China with a heartfelt wish: “I hope we can all love each other”. 🏓❤️
In addition to the SAS paper, we’re releasing a dataset of table tennis ball throws, and a simple MuJoCo simulation environment able to replicate trajectories from the real world, with data on specific serves and rallies. Find it on @Github ↓
With the SAS prompt, we can now use language models like Gemini to learn from a robot's history.
This allows the model to analyze parameter effects ⚡, and suggest ways to improve - similar to a real-life table tennis coach. 💡
We’re helping robots self-improve with the power of LLMs. 🤖
Introducing the Summarize, Analyze, Synthesize (SAS) prompt, which analyzes how they perform tasks based on previous actions and then suggests ways for them to get better using the medium of table tennis. 🏓
Our robot first trains in a simulated environment, which can model the physics of table tennis matches accurately.
Once deployed to the real world, it collects data on its performance against humans to refine its skills back in simulation - creating a continuous feedback loop.