Español
Português
Việt Nam
日本語
русский 
عربي 
Deutsch
한국어
Türkçe
Français
Italiano
The powerful functions of ChatGPT make many technology companies want to explore its application boundaries, such as giving it a robot body, what kind of surprises can it bring us?
A few days ago, Microsoft published a technical article on its official website, introducing that Microsoft is trying to let ChatGPT break through the limitations of computers to control robots, so that it can play a role freely in the real world.
According to the content of the article, Microsoft will extend the capabilities of ChatGPT to the field of robotics, and use language to intuitively control multiple platforms, such as robotic arms, drones and home assistant robots.
For example, if you say "please heat up my lunch" to a home robot, then the home robot will find the microwave to heat up the meal by itself. This is obviously our ideal robot, which can tell the robot what to do in our own words, just like we It's as easy as communicating with other people.
Language is the simplest and intuitive way for us to express our intentions, but now we can only rely on writing code to control the robot, which undoubtedly makes the operation much more complicated, and the Microsoft team is currently exploring how to change this reality. Try ChatGPT, OpenAI's new AI language model to enable natural human-computer interaction.
Application of ChatGPT on robots
ChatGPT is a language model trained on a large corpus of text and human interaction to be able to generate coherent and grammatically correct responses to a wide range of prompts and questions. To use ChatGPT to control robots to complete tasks, it is necessary to make it think beyond text , to deduce the physical world.
Microsoft researchers said: "The goal of the research is to see if ChatGPT can go beyond the scope of generating text and reasoning about real-world conditions to help robots complete tasks." Microsoft hopes to help people interact with robots more easily without having to learn complex programming language or detailed information about the robot system.
The current basis of robot use is engineers or technical users who can translate task requirements into system code. Engineers need to frequently write new codes and specifications to correct robot behavior. This process is slow (users need to write low-level code), expensive (requires highly skilled users with deep knowledge of robotics) and inefficient (multiple interactions required to get things working).
ChatGPT unlocks a new robot paradigm that can quickly convert natural language into code, so a non-technical person can also provide advanced feedback for a large language model (LLM) while monitoring the performance of the robot. Based on this, Microsoft has no In any case of fine-tuning, the knowledge of LLM is used to control robots of different shapes to complete various tasks.
It is understood that the current experiment has been able to input commands to the ChatGPT dialog box to let it control the robot to find "healthy drinks", "things with sugar and red signs" (in this case, a can of Coke) in the room. , and a mirror for drone selfies.
As it turns out, ChatGPT can indeed do a lot on its own, but some technical improvements are needed, including but not limited to special prompt structures, advanced APIs, and human feedback via text. Microsoft believes their current efforts are just a change in the way they develop bot systems. started, and hope that more researchers enter this exciting field.