武汉一群“00后”教机器人“做人”

Core Insights - The article highlights the emerging role of data collectors in the humanoid robot industry, emphasizing their importance in teaching robots to perform tasks through hands-on training and data collection [1][6]. Group 1: Role of Data Collectors - Data collectors, like Qi Dan, are responsible for teaching humanoid robots to perform various actions, providing essential data for the robots' learning process [2][3]. - The average age of data collectors at the company is 21, indicating a trend of attracting young talent into the industry [1][6]. - Each task performed by the data collectors generates significant data, with a single cleaning task producing over 4000 megabytes of data [2]. Group 2: Training and Data Annotation - The training process for robots involves breaking down actions into manageable steps, which can take two to three months for seemingly simple tasks [5]. - Data annotators play a crucial role in helping robots understand the context of their actions by providing detailed explanations and annotations for the data collected [4][3]. - The annotation process is time-consuming, with a five-minute video requiring approximately 40 minutes to annotate correctly [4]. Group 3: Industry Development and Talent Supply - The humanoid robot industry is rapidly evolving, with increasing specialization in roles such as data collectors and annotators [6]. - The Hubei Humanoid Robot Innovation Center collaborates with local universities to train students, creating a pipeline of talent for the industry [6]. - The center is noted for its extensive range of training scenarios, housing over 100 types of robots and generating 24,000 effective data entries daily [6].