正文 · AI 翻译
ITHome reported on October 19 that an AI lab focused on collecting real-world data plans to work with homeowners to record skilled tradespeople at work for use in training AI robots. The company, called HUMXN, says it will subsidize plumbing, electrical, HVAC, air duct cleaning and carpet cleaning services: residents who need them can receive these services free of charge in exchange for allowing personnel wearing “wearable data collection devices” into their homes to record the work.
HUMXN founder Thomas Unise said: “The next generation of AI will need more than just text and images from the internet. Robots need to understand how professionals interact with the physical world. Home service technicians perform millions of highly skilled tasks every day. We built this model so homeowners and service businesses can earn a financial return from the value of those professional skills.”
He added: “Millions of skilled tradespeople perform extremely complex tasks every day, but very little of that expertise exists in a standardized data format that AI systems can learn from directly. We want to build infrastructure that connects this knowledge with companies developing the next generation of intelligent devices, while ensuring that the people who create the value share in the proceeds.”
HUMXN says it is difficult to capture skilled tradespeople’s natural movements and behavior in a laboratory. By working with homeowners, it can collect data from different homes, types of equipment, fault scenarios, individual technicians and environmental conditions. These data can allow AI models to learn the full process by which technicians handle a task: inspecting the site, testing systems, choosing the right tools, responding to unexpected situations and determining whether the work meets standards.
The lab says the arrangement lets homeowners receive eligible home services for free while ensuring service personnel are paid as usual. In essence, HUMXN pays residents for permission to record video inside their homes and collect real-world training data. The program will launch first in 3 cities—Minneapolis, Chicago and Miami—to obtain data from different climates. If you live in one of these 3 cities and need a repair or service at home, you can visit HUMXN’s free home services webpage to see whether you qualify to participate.
ITHome notes that several AI labs, including Meta, Nvidia and OpenAI, have been accused of scraping data without permission to train models. By comparison, HUMXN’s model of paying directly for training data is fairly novel. However, the next generation of AI will probably be unable to train solely on text, images and videos scraped from the internet, and companies like these may have to pay for data generated in real-world settings.