> For the complete documentation index, see [llms.txt](https://docs.natix.network/whitepaper/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.natix.network/whitepaper/introduction.md).

# Introduction

Artificial intelligence is moving into the physical world. Autonomous vehicles, robotics, and embodied systems must operate in dynamic environments, requiring them to perceive, interpret, and respond to real-world conditions. As a result, the role of data is shifting from static datasets to continuous streams of real-world visual information.

This creates a new bottleneck. Training and validating Physical AI systems requires large-scale, diverse, real-world data, including rare and unpredictable edge cases. Today, this data is expensive to collect, limited in scope, and controlled by a small number of centralized players.\
\
At the same time, billions of camera-equipped devices already exist in the world, embedded in smartphones, vehicles, and other connected systems. NATIX builds on this foundation by enabling a distributed approach to data collection, transforming existing hardware into a scalable source of real-world visual data for Physical AI. 

<br>