Microsoft researchers program a robot to create the company’s logo using ChatGPT (Image: Microsoft)
OpenAI’s ChatGPT has already proven itself by generating essays that look like they were written by humans and basic programming. But Microsoft has found that the AI chatbot can also play a role in human-to-robot interactions. Most of the code written to control robots today is handwritten, and Microsoft wants to change that. The company is experimenting with using ChatGPT as an intermediary between humans and helper robots.
In a blog post, Microsoft points out the difficulties associated with traditional robot programming. Current robotics pipelines begin with an engineer translating the requirements of a task into code for the system. The engineer writes code and then rewrites it to fix a bot’s behavior until it’s perfected. The process is lengthy, expensive and inefficient.
ChatGPT can be used to greatly simplify this process, allowing even non-technical users to program robots. High-level feedback (commands in plain English) can be submitted to ChatGPT, which then generates code, mostly in Python, to drive the robot accordingly. “Without fine-tuning, we use the knowledge of the LLM to drive different robot form factors for a variety of tasks,” adds the blog post.
As a demonstration of this technique, Microsoft shared a video in which ChatGPT wrote code to control a real drone using simple user prompts. While the objects the robot interacted with were predefined for ChatGPT along with their location data, the results were nonetheless remarkable.
For example, type “I’m still thirsty. Can you find something with sugar and a red logo?” tricked ChatGPT into understanding that the user was referring to the can of coke in the room. The AI chatbot then programmed the bot to navigate through the room towards the can.
The researchers also noted in their blog post that “ChatGPT asked clarifying questions when the user’s instructions were ambiguous and wrote complex code structures for the drone, such as: B. a zigzag pattern to visually inspect shelves.”
ChatGPT can write code because it was trained on huge amounts of data that also contained code. The chatbot has already proven its ability to solve coding problems while providing detailed explanations for them – and also seeking clarifications when needed.
Aside from the drone experiment, Microsoft researchers also tested ChatGPT to use a robotic arm to move around blocks to create the Microsoft logo. They also tested whether ChatGPT is able to decide where to go for the robot based on real-time sensor feedback.
© IE Online Media Services Pvt Ltd
