The Menteebot prototype, which showcases the incorporation of artificial intelligence at all levels of its operation, was unveiled today. This prototype utilizes a unique approach known as Sim2Real, where machine learning takes place in a simulated environment before being adapted to the real world with minimal data requirements.
The robot can map its environment in real-time using algorithms based on NeRF technology, creating cognitive maps that allow for efficient navigation and obstacle avoidance. Additionally, large transformation-based language models are used to interpret commands and coordinate complex tasks that require a combination of leg movement and manual work.
The end-to-end action demonstrated by the robot includes responding to verbal commands, navigating the environment, identifying objects, and completing tasks using natural language understanding. The final version of the robot is expected to go into production in 2025, operated solely by camera-based sensing and electric motors developed by the company. These technologies enable the robot to perform tasks, engage in conversations, and continuously learn new skills.
Professor Amnon Shashua, the founder and chairman of the company, highlights the convergence of various technologies in the field of AI, leading to the development of robots that can navigate human environments and perform a wide range of tasks. This advancement has the potential to revolutionize the way robots interact with and assist humans in various settings.