In a groundbreaking development, 1X Technologies, a Norwegian-based robotics company supported by OpenAI, recently unveiled the remarkable capabilities of its humanoid robot, EVE. This android, equipped with neural networks, is now capable of learning and mimicking human actions by watching videos. Unlike traditional programming methods, EVE’s autonomy is achieved through a single vision-based neural network operating at 10Hz, controlling various functions such as driving, arms, gripper, torso, and head without the need for intricate coding.
What sets EVE apart is the demonstration of learning in real-world scenarios. The company ensures the safety credentials of its androids by testing each one extensively before deployment, emphasizing their commitment to working alongside human counterparts to meet global labor demands. Notably, EVE’s actions are not controlled by graphics, video speedups, or teleoperation; instead, everything is orchestrated through neural networks, operating at 1x speed.
1X Technologies utilized thirty EVE robots to compile a diverse dataset showcasing various human-like behaviors. This dataset was then used to train a base model capable of identifying standard human tasks, such as tidying homes, picking up objects, and interacting with others. The model’s adaptability was further demonstrated by refining specific capabilities, such as opening different types of doors, with the ability to onboard additional skills within minutes. This approach marks a shift towards a new era of “Software 2.0 Engineers,” as 1X celebrates the next generation of operators who advance robotics through data, rather than traditional coding methods.
This breakthrough signifies a monumental leap in the field of robotics, where the ability to teach digital employees is no longer constrained by the availability of AI engineers. 1X Technologies envisions a future where the flexibility and choices to meet customer demands are enhanced, thanks to the innovative capabilities of their intelligent agents, paving the way for a new era in human-robot collaboration.
Key Highlights:
- EVE’s Breakthrough Learning Abilities: 1X Technologies announces a significant breakthrough in teaching skills to its humanoid robot, EVE. This android can now learn and replicate human actions by watching videos, thanks to specialized neural networks.
- Real-world Testing for Safety: The Norway-based robotics company ensures the safety credentials of its androids by extensively testing them in real-world scenarios before deployment, emphasizing their commitment to working alongside human labor.
- Neural Network Autonomy: EVE’s actions are controlled by a single vision-based neural network operating at 10Hz. This neural network oversees various functions, including driving, arms, gripper, torso, and head, without relying on traditional programming methods or external controls.
- Data-driven Training Process: 1X Technologies uses thirty EVE robots to compile a diverse dataset showcasing human-like behaviors. This data is then used to train a base model capable of identifying and replicating standard human tasks, demonstrating the adaptability and efficiency of their robotic systems.
- Shift Towards “Software 2.0 Engineers”: The company celebrates a new generation of operators who advance robotics through data, not traditional coding methods. This approach allows for the rapid onboarding of additional skills within minutes, marking a transformative shift in teaching and evolving digital employees.
- Flexibility in Meeting Customer Demand: 1X Technologies envisions a future where teaching robots is no longer constrained by the availability of AI engineers. This breakthrough promises flexibility and choices in meeting customer demands, representing a paradigm shift in the landscape of human-robot collaboration.
Reference:
https://readwrite.com/1x-robotics-company-showcases-its-androids-driven-by-neural-networks/
