Microsoft wants to use ChatGPT to control robots – will a future like the movie “Terminator” come?
- Tram Ho
On October 26, 1984, Orion Pictures shocked the movie world when it introduced The Terminator, a movie about artificial intelligence robots released by a technology company. Coincidentally, nearly 40 years later, it is widely reported that technology giant Microsoft is preparing to do the same.
As part of increasing cooperation with OpenAI, Microsoft is planning to use its ChatGPT chatbot to control robots.
In its blog post, Microsoft said: “Our goal in this study was to see if ChatGPT could think beyond the lines of text and make inferences about the real world to perform tasks. We want to make it easier for people to interact with robots, without having to learn complex programming languages or the details of robotic systems.”
Robots now run our world – from robotic arms that make products in factories to robot vacuums in our homes. But every time you want a robot to do something new, you need an engineer with a full understanding of the robot to write the code and test it over and over to make sure it works.
But imagine a world where you can talk directly to the robot and give it commands in your usual sentences, in any language. So what happens when robots understand the laws of physics – Microsoft thinks that’s good. However, a future like science fiction is far from becoming a reality.
In their report, Microsoft researchers say they plan to one day help these robots heat food at home when they hear a user say ” heat my lunch ” as well as do other housework.
The researchers even detailed the use of specific prompts to allow ChatGPT to control a small drone. By writing instructions in ChatGPT’s chat box, researchers were able to direct the drone to find objects in the room such as “a healthy drink”, “something with sugar and red logo” (probably a can of Coke).
Often, what makes programming robots so difficult is that computers communicate with mechanical parts in machine language – a language much more primitive than conventional programming languages. This in turn makes ChatGPT or today’s chatbots suitable for that job – they can translate natural language into machine language quickly.
Recently a clip went viral on Twitter showing how that could happen. A ChatGPT-integrated robot (claimed in the video) can understand its speaker’s questions and answer them – even eye movements give the robot facial expressions when it’s talking. In other words, for the first time, ChatGPT can communicate with people in the form of a human instead of through a computer interface as in the past.
Of course, it will be a long time before a robot with the capabilities of the Terminator appears, but it seems that Microsoft researchers are also very cautious about applying AI models in robot control:
“We emphasize that these tools should not be given full control of the robot, especially for applications where safety is important. Due to the tendency of LLM models to often produce unrealistically false responses , it is important to ensure the solution quality and the safety of the code lines with human supervision before executing it on the robot .”
An interesting thing is that this idea has also been applied by Google in its Everyday Robots project – AI-integrated robots that can become a housekeeper in each family, helping to clean, make coffee, close opening doors and other errands. Although it has been researched and officially established for more than a year, the Everyday Robots project was finally closed a few weeks ago during a storm of cost and staff cuts at Alphabet, the parent company of Google. Now Google’s unfinished dream is inherited by its own rival, Microsoft.
Source : Genk