The world of robotics is about to undergo a significant transformation, and the front lines of this revolution are in Silicon Valley. Meet Fernando Flores, a robot puppeteer who spends his days teaching humanoids everyday tasks like pouring coffee. It's a repetitive and challenging task, but Flores is on the cutting edge of an emerging industry that is preparing for the impending robot invasion.
The Rise of Robot Trainers
The tech industry's ambitious plans for humanoid robots are no longer a distant dream. Companies like Tesla, 1X Technologies, and Figure AI are gearing up for mass production, with optimistic projections of millions of robots hitting the market annually. These robots need an immense amount of data on human movement to function, and that's where companies like Encord step in.
Data Collection and Training
Initially, data collection involved humans wearing cameras to record their daily chores. However, this method lacked the precision required for robot hands to perform flawlessly. Now, the industry has shifted to direct human guidance, using expensive rigs to control robot movements. This approach provides richer insights into motor skills and object manipulation, with Encord charging up to $1,000 per hour for this training data.
The Challenge of Everyday Tasks
While robots have mastered walking, they still struggle with basic household tasks. Opening doors, fridges, and washing machines seamlessly is a challenge, and they lack the understanding of how to grasp handles or apply the right force. Flores has perfected the art of pouring coffee with robot arms, but spills are common, and he has to delete records of failed attempts.
Replicating Home Environments
Encord has created a test facility that replicates a standard American home, complete with a living room, kitchen, and bathroom. Here, pilots train robots to perform tasks like rearranging desks, opening refrigerators, and even turning water faucets on and off. It's a meticulous process, and the data collected is crucial for robots to learn and correct their mistakes independently.
Industrial vs. Home Use
Many robotic data companies are focusing on industrial use cases, as robots perform better in structured and predictable environments like factories and warehouses. Home tasks, on the other hand, are more varied and messy, making them a tougher challenge for robots.
The Role of Teleoperators
Teleoperation is emerging as a vital component of the humanoid robot business. Employing operators in countries with lower wages can significantly reduce costs. This month, a San Francisco-based robot cleaning service, Gatsby, completed a cleaning task in the U.S. using a teleoperator in Mexico. Encord plans to set up a teleoperations center, recognizing that even as robots become more sophisticated, they will still need human intervention at times.
The Future of Robotics
As the robotics industry expands, the demand for robot trainers and teleoperators will grow. The data collected from these human-robot interactions is crucial for the development and deployment of humanoid robots in various settings. The future of robotics is an exciting and rapidly evolving field, and the work of robot trainers like Flores is a critical part of this journey.
