HUMXN, an AI research lab, is offering free plumbing and HVAC repairs to homeowners in Minneapolis, Chicago, and Miami as part of an effort to collect real-world training data for robots and AI systems. The program is aimed at capturing how skilled technicians handle complex physical tasks in actual homes rather than in laboratory settings.
The company's founder, Thomas Unise, said the next generation of AI needs more than text and images pulled from the internet. He said robots must learn how skilled people interact with the physical world, and that home-service technicians perform millions of highly technical tasks every day. According to Unise, HUMXN is building a way for homeowners and service companies to benefit financially from the value of that expertise.
Unise also said that millions of skilled workers carry out complex tasks daily, but very little of that knowledge exists in a form AI systems can learn from. He described HUMXN's goal as building infrastructure that links that expertise to companies developing intelligent machines, while ensuring the people generating the value share in the economics.
HUMXN said capturing technicians' natural movements is difficult in a lab. By working with homeowners, the company said it can record data across different homes, equipment types, problems, technicians, and environments. The company said this would also show AI models how technicians approach a job, from inspecting the situation and testing systems to selecting tools, adjusting to changing conditions, and judging whether the work was finished correctly.
Under the arrangement, homeowners get qualifying work done on their property at no cost while service professionals are still paid. HUMXN effectively pays residents for permission to record on their premises and gather the training data. The three launch cities were chosen to capture data from places with very different climates. Residents in those cities can check whether they qualify on the HUMXN Free Home Service webpage.
The program stands out because several AI labs, including Meta, Nvidia, and OpenAI, have faced accusations of taking data to train their models. HUMXN's approach of paying for training data contrasts with those allegations. The company's effort also reflects a broader constraint facing AI developers: the next generation of systems likely cannot be trained on text, images, and video scraped from the internet alone, which may push companies to pay for data generated by real-world activity.
More AI news from TechManNews.
