Microsoft announced in a blog post that their local researchers conducted a test on ChatGPT using robot arms, drones, and home assistant robots. The respective company then announced that the experiment’s results were positive. Furthermore, the company has reportedly stated that the results indicate that the AI-generated tool has a bright future in the robotics realm.
Microsoft’s Strategy
The entrepreneur claims that the creation of ChatGPT has ushered in a new era of modern AI. This was supported by reports that, following the company’s announcement that it would be integrating an AI language model into Bing, several other technology companies, including Google, began introducing their own works centered on the technology.
In light of this, several Chinese companies have reported their own research and development of various ChatGPT-like projects. In the midst of new emerging projects similar to that tool, the company has stated that in order to remain competitive, they must move forward and apply it in other areas such as robotics.
The Experiment
According to the Microsoft Autonomous Systems and Robotics Research Group, the tool received various tasks on various platforms during their experiment. The experiment used various design principles, such as special prompting structures, high-level APIs, and text-based human feedback. The researchers then stated that, while the technology “still requires some assistance,” the results of the experiment demonstrated that “ChatGPT can do a lot on its own.”
The researchers also stated that by adhering to their set of design principles, the tool can generate code for robotics scenarios. They went on to say that they were able to use the large language model knowledge to control different robot form factors for a variety of tasks without the need for fine-tuning.
In addition to following text orders to take a selfie in front of a mirrored surface and investigate a shelf in a mowing pattern, ChatGPT was able to carry out commands from finding a drink, identifying a drink based on descriptions, and offering a “healthy option. Also, it passed the manipulation test by stacking blocks and using its knowledge base to build the company’s four-color logo.
Microsoft went on to say that although the experiment’s results were encouraging and favorable, they still only represented “a small portion” of what could be achieved when using massive language models on robots. Additionally, the business has acknowledged that ChatGPT is not yet completely prepared to assist robots in carrying out duties, and it has even urged other researchers and enthusiasts to “always take the required safety precautions.”
