Home Latest Feeds Technology News Microsoft tested ChatGPT on robots, the results are on video

Microsoft tested ChatGPT on robots, the results are on video

Microsoft tested ChatGPT on robots, the results are on video


Microsoft researchers recently tested ChatGPT on robotic arms, drones, and home assistant robots. The company said that the overall conclusion of the experiment produced excellent results, indicating a bright future for ChatGPT in the field of robotics.

The appearance of ChatGPT marked the birth of the modern AI era globally. After Microsoft announced that it would integrate the AI ​​language model into Bing, other companies such as Google began sharing their own work around the technology. A handful of Chinese companies have also been reported to be researching and developing their own ChatGPT-like projects. With this, Microsoft realized that the only way to stay on top was to move forward and apply ChatGPT to other areas, such as robotics.

In the work shared by the Microsoft Autonomous Systems and Robotics Research Group, ChatGPT was given different tasks using different platforms. The tests also included various design principles, including special prompt structures, high-level APIs, and human feedback via texts. The group reported that while the technology “still needs some help,” the results of the project proved that “ChatGPT can do a lot on its own.”

“Following our design principles, ChatGPT can generate code for robotics scenarios. Without any fine-tuning, we take advantage of the LLM (large language model) knowledge to associate the shapes of different robots with different tasks,” the group shared.

In one of the tests carried out in the project, ChatGPT was called upon to control a drone and the Microsoft AirSim simulator. In some of the videos shared, ChatGPT was able to perform commands ranging from searching for a drink, identifying the drink by its description, and offering a “healthy option.” It also successfully followed a text command to take a selfie in front of a reflective surface and examine a shelf following a lawnmower pattern. When used in a simulated industrial control scenario, the researchers reported favorable results in the aerial obstacle avoidance test, adding that it “was able to effectively analyze the user’s high-level intent and geometric cues to accurately control the drone.”

In a more incredible scenario, ChatGPT passed the manipulation test by stacking blocks and using its knowledge base to create the four-color Microsoft logo.

“We used conversational feedback to teach the model how to assemble the initially given APIs into more complex, high-level functionality: this was coded by ChatGPT itself. …The model showed an impressive example of bridging the textual and physical domains when tasked with building the Microsoft logo from wooden blocks, he was able not only to recall the logo from his internal knowledge base, but also to draw the logo (as SVG code) and then use the skills learned above to figure out which of the existing robot actions its physical form can be assembled,” explained the group.

Although the results of the project look promising, Microsoft emphasized that the work is “only a small fraction” of what can be achieved by applying large-scale language models to robots. In addition, the company reminded that ChatGPT is not yet fully ready to help bots perform tasks and warned fans and other researchers to “always take the necessary safety precautions,” reported MS Poweruser.

Hardware, software, tests, interesting and colorful news from the world of IT by clicking here!



Please enter your comment!
Please enter your name here