Data for the Real World
Plummeting sensor costs make dense physical-world data collection feasible for AI models
“More real-world data enables precise modeling, and once you can model a system, you can control it.”
Y Combinator is signaling investment interest in startups collecting dense physical-world data for industries like energy, agriculture, logistics, and construction — sectors still running on sparse sensor data and intuition-based models. The thesis is that improving foundation models plus falling sensor costs now make it viable to build AI-powered predictive and control systems for the physical world. Examples cited include Gecko Robotics and Source Air, with ambitions as large as steering hurricanes and reversing desertification.