Microsoft has announced its investigation into human-robot interactions using the ChatGPT artificial intelligence technology. Developed by OpenAI, ChatGPT can generate ideas, link concepts, and recall conversations. Over the past month, Microsoft has been integrating this technology into its various services, starting with Azure OpenAI Service in mid-January, and later expanding it to Bing search engine and Edge browser. The latest development includes a preview of the new Bing AI-based feature for mobile apps, Edge browser, and Skype search engine. Microsoft revealed that the technology is now being tested on robots to enable natural interactions with humans.
The company aims to explore ChatGPT’s ability to reason beyond text and apply it to robotics tasks to facilitate easier human-robot interactions without the need for complex programming languages. The goal is to teach the chatbot to solve problems considering factors such as environmental context and the impact of physical actions on the world. OpenAI’s product has autonomous capabilities but also incorporates design principles to guide language models in robotics tasks. Microsoft envisions ChatGPT revolutionizing robotics by streamlining the typically slow, costly, and inefficient process of rewriting robot behavior code.
To enhance the Language Learning Model (LLM), Microsoft has established design principles for robotic task instructions, defining a set of Robotics Application Programming Interfaces (API). Developers provide task descriptions to ChatGPT, explicitly instructing the chatbot on task objectives, constraints, and response guidelines. By evaluating the AI’s responses for quality and safety, engineers can refine the instructions iteratively until the desired outcome is achieved and implement the final AI chat code onto the robot.
Microsoft demonstrated experiments showcasing robots’ responses using the OpenAI-developed model. From controlling a real drone to coding complex patterns like zigzag shelf displays and analyzing precise control signals in Microsoft AirSim simulator scenarios, ChatGPT exhibited adaptive skills. It even engaged in existential crises and manipulation tasks with a robotic arm, stacking wooden blocks based on learned capabilities and navigating objectives such as object search tasks.
The experiments showcased the potential of leveraging ChatGPT in robotic applications, enabling robots to carry out various tasks and adapt to new challenges efficiently. Microsoft’s innovative approach highlights the future possibilities of human-robot interactions and the integration of AI technologies in enhancing robotics functionalities.
About the Author
