TechCrunch AIJun 17, 2026, 3:00 PMTim Fernholz

Robot Training Data Is Dirty, Unglamorous Work — AI Labs Are Already Paying XDOF to Do It

Original: Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it.

AI labs are outsourcing the messy, labor-intensive work of collecting robot training data as physical AI scales up.

Physical AI systems need vast amounts of real-world demonstration data to approach LLM-level capability, but gathering it requires human operators physically performing tasks — work that can't be scraped from the internet. Unlike text data, robot training data demands presence, equipment, and repetitive labor. Some AI labs are already turning to paid data-collection pipelines, including XDOF, to meet this growing operational need.

The race to build capable physical AI — robots and embodied systems that act intelligently in the real world — has run into a fundamental bottleneck: training data. Unlike large language models, which were trained on trillions of tokens harvested from the open web, robots need something far harder to acquire: grounded, physical demonstrations of real-world tasks. This TechCrunch report, published June 17, 2026, examines how AI labs are confronting a data problem that is as much a human-labor and logistics challenge as it is a technical one.

Full summary

Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.

See Pro plans →

Want the original English / full article?

Read on TechCrunch AI →

Summaries are AI-generated; the original article is authoritative.