Mentee Robotics is advancing the development of a humanoid robot designed to perform a broad range of tasks in homes and industrial settings. This robot demonstrates a complete operational cycle from receiving a verbal command to executing complex actions.
The company, founded in 2022 by AI experts Prof. Amnon Shashua, Prof. Lior Wolf, and Prof. Shai Shalev-Shwartz, has worked in stealth for two years to create this sophisticated technology. The trio's expertise spans AI, computer vision, and machine learning.
The newly introduced Menteebot prototype integrates AI across all operational facets. It utilizes a Sim2Real machine learning approach for movement, training in a virtual environment before adapting to real-world conditions with minimal data. Environmental mapping is performed in real-time using NeRF-based algorithms, which generate 3D cognitive maps that store semantic data, aiding the robot in navigating and locating objects.
For command interpretation and task planning, the robot employs Transformer-based Large Language Models (LLMs), allowing it to process instructions and strategize actions effectively. The design also focuses on the robot's mobility and dexterity, ensuring it can dynamically balance while carrying weights or extending its limbs.
Prof. Amnon Shashua, chairman of Mentee Robotics, highlighted the significance of this development. "We are on the cusp of a convergence of computer vision, natural language understanding, strong and detailed simulators, and methodologies for transitioning from simulation to the real world. At Mentee Robotics, we view this convergence as the foundation for designing the future general-purpose bipedal robot that can move and operate as humans do, performing household tasks and learning new tasks through imitation," said Shashua.
The production-ready version of the robot is anticipated for deployment by Q1 2025. It will feature camera-only sensing, proprietary electric motors for enhanced dexterity, and fully integrated AI capabilities, enabling it to perform complex reasoning, engage in conversations, and learn new tasks on the fly.
.png)
