OpenAI-backed robotics firm 1X has unveiled a captivating video showcasing a fleet of wheeled service robots effortlessly transitioning between tasks as they efficiently clean and organize an office space, expertly guided by a voice-controlled natural language interface.
Founded in 2014, Halodi Robotics aimed to design and develop robots that seamlessly integrated into everyday office environments, working collaboratively with humans. Headquartered in Norway initially, the corporation subsequently established a second base of operations in California in 2019. It was during this period that we first encountered a pre-production prototype of a wheeled humanoid robot dubbed ?
Halodi’s growth culminated in 2022 with a mission “to harmonize robotics and AI, laying the foundation for embodied learning.” The company has already developed a bipedal robot and human-like hands; however, most of its current focus is on creating helpful assistants that can navigate office spaces. These bots will be capable of perceiving natural language and physical environments, enabling them to perform actual tasks throughout your office and world.
One X now investigates whether a pure language interface has been developed, allowing an operator to control multiple humanoids using voice commands, with the robots seamlessly integrating a sequence of executed tasks to complete complex tasks.
Voice Instructions & Chaining Duties | 1X AI Replace
The company claimed to have successfully developed an autonomous mannequin capable of assimilating various tasks into a single behavioral AI model, encompassing responsibilities such as retrieving items from a shopping bag and determining their placement, cleaning up spills, and folding garments.
The notion is well-known that optimizing the performance of a single task within a relatively small multitasking framework can have unintended consequences on the behaviors of other tasks within that framework, potentially leading to suboptimal outcomes. While increasing the model’s parameters might speed up training, this would likely come at the cost of longer training times and potentially slower learning rates.
By integrating a voice-controlled natural language interface, operators can seamlessly combine short-term capabilities across multiple small models to create longer, more sophisticated ones. This enables them to merge single-task models into goal-conditioned models as progress marches towards a unified model capable of autonomously executing complex actions with AI-driven precision.
“Through our intuitive high-level language interface, users gain unparalleled insights into data collection, revolutionizing their experience and capabilities,” said Eric Jang of the company in a recent blog post. By leveraging virtual reality, operators can command multiple robots using high-level commands, allowing low-level control algorithms to execute precise actions that align with those directives. As a consequence of high-level actions being dispatched from time to time, operators are occasionally able to manage robots remotely.
The Eve humanoids featured in the video, typically, do not rely on tele-operation; instead, their actions are orchestrated by a complex neural network, allowing for autonomous decision-making and execution.
Supply:
