This is the SAS Prompt: a new method for robot self-improvement using large language models. It involves using an LLM to generate policy parameters for a robot. The trace of the robot’s actions are sent to the LLM for improvement. This loop is repeated until the robot can execute tasks. The SAS Prompt stands for summarize, analyze, and synthesize.
Hot Computer Stuff
SAS Prompt: Large Language Models as Numerical Optimizers for Robot Self-Improvement
By using this approach, new better parameters are synthesized using current data via AI. That enables robots to learn a wide variety of skills. For this project, a table tennis robot was improved with AI. It learned to hit the robot to specific parts of a ping pong table.
[HT]
**Disclaimer: Our posts may feature aff links. As an Amazon Associate, we earn from qualifying purchases. Please read our disclaimer on how we fund this site.
