Why a Robot Trained in Texas Might Struggle in Tokyo.
Physical AI is attracting extraordinary investment, with robotics and humanoid systems moving from research labs into real-world environments. But as developers race to build more capable, general-purpose robots, an important challenge is emerging:
Does the training data reflect the diversity of the world these machines are expected to operate in?
A robot trained to navigate a warehouse, handle household objects or interact with people is learning from specific environments, behaviors and routines. What works in one country may not translate seamlessly to another.
From the layout of a kitchen to the way people interact with machines, cultural and geographic differences introduce complexity that model architecture and computing power alone cannot solve.
In a new Forbes Technology Council article, Robots Don’t Have a Culture Problem: Their Training Data Does, DATAmundi CEO Véronique Özkaya explores why culturally representative human data is becoming critical to the commercial success of physical AI.
The article examines the growing gap between investment and real-world deployment, where value is likely to emerge across the physical AI ecosystem, and why global data collection infrastructure could become a competitive advantage for frontier AI labs and robotics developers.
Read the full article on Forbes → Robots Don’t Have A Culture Problem, But Their Training Data Does
