Microsoft has demonstrated the capabilities of ChatGPT, the conversational AI developed by OpenAI, to control a series of robotic tasks with just human language instructions. In a groundbreaking experiment, ChatGPT converted these instructions into code to facilitate communication between humans and machines without any particular knowledge of computer science. One of the tasks included controlling a drone for shelf inspection.
Expanding ChatGPT Capabilities
Microsoft aimed to expand ChatGPT’s abilities to robotics. Through highly precise requests, the artificial intelligence was able to produce code to control various robots, including a drone.
The Challenge
Although ChatGPT is capable of generating code, controlling a drone required the consideration of various factors, such as physics laws and environmental constraints. Consequently, the requests had to contain an enormous amount of precise information to generate workable code without posing any risk to the drone or its surroundings.
The Successful Outcome
Ultimately, the drone flew correctly, maneuvered through several obstacles, and even captured a photo in flight on demand. The tests were conducted using the Microsoft AirSim platform, and the company achieved similar success in testing other robots.
Revolutionizing Human-Robot Interactions: Drone Control
Although still experimental, the idea is to enable the control of machines with natural language instructions, while an AI like ChatGPT converts those requests into code. With ChatGPT (or any other AI), it will no longer be necessary to be an expert in coding to take control of a robot or drone. Microsoft has already integrated ChatGPT into its Bing search engine and plans to do the same with Outlook and the entire Office suite.