According to Peopletalk, an international team of researchers has developed a new technology that enables effective cooperation between humans and robots. The new system allows robots to detect their own mistakes and correct them in real time.
The project is aimed at simplifying interaction in industrial and manufacturing environments. It is based on Retrieval-Augmented Generation (RAG) technology and specially “finely tuned transformer neural networks.”
The main innovation is the regret-based learning mechanism. Through this method, the robot learns from its own mistakes and performs subsequent actions in the right direction.
Experts note that such a system reduces the number of errors in the production process, increases efficiency, and creates a comfortable working environment for humans.
Read 'Zamin' on Telegram!
